- Email: [email protected]
Route planning evaluation of a prototype optimised infield route planner for neutral material flow agricultural operations
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/issn/15375110
Research Paper
Route planning evaluation of a prototype optimised infield route planner for neutral material flow agricultural operations Gareth T.C. Edwards a,*, Jørgen Hinge b, Nick Skou-Nielsen a, Andres Villa-Henriksen a, Claus Aage Grøn Sørensen c, Ole Green a a
Agro Intelligence, Agro Food Park 13, 8200 Aarhus N, Denmark AgroTech A/S, Agro Food Park 15, 8200 Aarhus N, Denmark c Aarhus University, Nordre Ringgade 1, 8000 Aarhus C, Denmark b
article info
The need to decrease unit production costs has led agricultural industries to develop larger
Article history:
and consequently heavier machinery. While this has increased the productivity of single
Received 1 July 2016
machines, it has also caused significant soil compaction, which may cause reduced crop yield
Received in revised form
and crop quality. Therefore, mechanisation solutions that have both lower unit costs and
7 September 2016
reduce the risk of soil compaction are needed. Optimising infield routes will reduce labour
Accepted 11 October 2016
costs, fuel consumption and field trafficking intensity, providing important benefits for
Published online 11 December 2016
infield operations. In this paper, a prototype of an optimised infield route planning tool for neutral material flow operations is evaluated. The evaluation parameters focused on dis-
Keywords:
tance and traffic intensity reductions, comparing the routes proposed by the tool prototype
Optimised infield route planner
and the routes followed by a professional operator during mowing operations. The tool re-
Path planning
quires some minimum inputs: field boundaries, field gates, working width and minimum
Performance evaluation
turning radius, in order to provide an optimised route. Twelve fields were recorded by a Global
Travelled distance reduction
Positioning System (GPS) during mowing operations and later compared with the routes
Trafficking intensity
proposed by the tool. In all fields, the operator's normal route was longer in distance than the
Precision agriculture
route proposed by the tool, being up to 18.4% longer. In total, 9.2 km of infield distance was saved, i.e. 7.5%. The traffic intensity was reduced in all fields, except for two of the smallest fields, where it equalled that of the normal route. Specifically, the traffic intensity was reduced in the working areas, as the tool confined all non-working distance to the headlands. © 2016 The Authors. Published by Elsevier Ltd on behalf of IAgrE. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.
Introduction
The increasing world demand on agricultural products has led the farming industry to increase productivity by diverse
solutions from different disciplines, e.g. genetics, agronomy or engineering (Tilman, Cassman, Matson, Naylor, & Polasky, 2002). In the last decades, the engineering focus has been to develop large, powerful, and high-capacity machinery, in order to decrease unit costs. However, this development has
* Corresponding author. E-mail address: [email protected] (G.T.C. Edwards). http://dx.doi.org/10.1016/j.biosystemseng.2016.10.007 1537-5110/© 2016 The Authors. Published by Elsevier Ltd on behalf of IAgrE. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
150
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Nomenclature GPS ORP NMF Ww Rm Fg Fb Ph Pr Pc
Global Positioning System Optimised Infield Route Planner Neutral Material Flow Working width Minimum turning radius Field gates position Field boundaries position Headland paths Row paths Connection paths
made machinery heavier, and it is consequently creating important subsoil compaction problems, which may result in lower yields and reduced fertiliser efficiency, along with higher water-logging, run-off and erosion problems (Hamza & Anderson, 2005). Soil degradation is an increasing problem worldwide, mainly caused in modern agriculture by subsoil compaction caused by heavy machinery (Hamza & Anderson, 2005). Ac la, H cording to Kroulı´k, Kumha ula, and Honzı´k (2009), p. 95% of the area in a field in conventional agriculture is run over at least once a year, meaning that the problem is generalised throughout the field. Repeated traffic over an area highly increases the risk (Keller, Arvidsson, Dawidowski, & Koolen, 2004). Soil compaction means increased bulk density and homogenisation of the soil, causing decreased aeration and water infiltration and increased penetration resistance, which impedes a proper root development and limits the biological ska_ activity of the soil (Gła˛b, 2014; Horn, Domz_ zał, Słowin Jurkiewicz, & van Ouwerkerk, 1995). Subsoil compaction has long term effects on the soil which are difficult to solve, therefore the most effective practice is to avoid or reduce the compaction as much as possible, rather than apply post-effect solutions, e.g. deep ripping (Laura Alakukku, 1996). Furthermore, soil compaction also has effects on greenhouse gas emissions. For example, Bhandral, Saggar, Bolan, and Hedley (2007) found in grassland soils that the N2O emissions were between 2 and 14 times higher for a compacted soil than for an uncompacted soil, with the rate for nitrate fertilisation being especially high. These results were corroborated by Uchida, Clough, Kelliher, and Sherlock (2008). Regardless of compaction problems, farmers need to decrease unit costs in order to adapt to and compete in the globalised modern market system, where apart from the exceptional increases in 2007e08, and 2011, low prices dictate their agenda (EU, 2016; FAO, 2015). There is therefore a need for solutions that can reduce both the unit costs and reduce the risk of soil compaction. Computer based tools can both optimise farming operations, as well as minimise the risks from soil compaction, making the whole system more sustainable. Although computer innovation in farming has been more common for business related activities than specifically for farming (Lewis, 1998), in the last decade the number of computer based tools in precision farming has grown considerably (Kaivosoja, Jackenkroll, Linkolehto, Weis, & Gerhards, 2014). One of these tools is an infield route planner, which optimises the route followed by the
machinery. Optimised infield route planning can reduce labour costs and fuel consumption, in addition to reduce field trafficking, which is one of several solutions proposed to reduce soil compaction (Bochtis, Sørensen, Busato, & Berruto, 2013; Hamza & Anderson, 2005). Moreover, infield route planning can be used for controlled traffic farming, which offers farmers an opportunity to not only reduce soil compaction, but also to restore their soils (McHugh, Tullberg, & Freebairn, 2009). Optimised route planners are already commonly used in Global Positioning System (GPS) based applications in mobiles and computer systems, for personal use, as well as for industrial and logistical uses; however they are still in the development stages for farming applications (Sørensen & Bochtis, 2010). Several studies and projects are working with optimised algorithms and solutions for route planning in precision agriculture, and especially for the emerging development of field robots. Different activities require different approaches, as some activities have capacity constraints and may require aid from support units (Bochtis, Sørensen, & Vougioukas, 2010), e.g. cistern trucks for replenishing fertilisers or pesticides. Furthermore, there may be in-field and inter-field routes that need to be optimised, making route planning in agriculture a challenging task, as there are infield attributes (e.g. working tracks and headland passes), and inter-field configurations (e.g. field gates and road networks) to be considered (Jensen, Bochtis, Sorensen, Blas, & Lykkegaard, 2012). However, none of the above mentioned models have been adapted into functional tools which can be used by a tractor drive to optimise in field operations in realtime. An optimised Infield Route Planner (ORP) prototype tool is evaluated. The ORP tool is designed for ‘neutral material flow’ (NMF) field operations, i.e. operations where there is no flow of material into or out of the field (e.g. tillage, cultivation, mowing) (Bochtis & Sørensen, 2009). Here, the tool is evaluated by comparing the distance travelled by a professional operator during mowing operations, with the optimised distance proposed by the ORP tool. Optimised route planners can save travelling distance inherent in agricultural operations, reducing consequently unit costs, and compaction risks. Secondarily, the ORP tool is evaluated by comparing the trafficking intensity as derived from the recorded data and the modelled optimised route, in order to assess the potential efficacy of the tool for reducing soil compaction.
2.
Material and methods
Applied mathematical modelling provides many possibilities to improve the effectiveness of precision agriculture. In this case, the driving pattern of agricultural machinery in the field is optimised by the application of an ORP prototype tool. The developed ORP tool has been evaluated by comparing the routes followed by an experienced tractor driver during mowing operations and the ones proposed by the ORP. The tool and the conditions are explained below.
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
2.1.
Description of the ORP tool
The Optimised Infield Route Planner (ORP) tool is designed as an application software and was developed by Agro Intelligence Aps, Aarhus, Denmark. The application creates routes for machines to follow when executing infield operations and with the aim of minimising the distance travelled by the machine. For this, the developed application requires the following inputs: field boundaries position (Fb), field gates position (Fg), working width (including any possible overlap) (Ww), and minimum turning radius of the machine (Rm). Once the inputs are given, the field is automatically divided into headlands and working rows, with their corresponding headland paths (Ph), row paths (Pr) and connection paths (Pc). The field boundaries are automatically registered as the outer edge of the working width during the operator's driving through the first track in the headland. The gates are also registered as the entry point of the machine in the field. The inner edge of the first headland defines the second and subsequent headlands, as well as the main working area. The number of headlands is calculated according to the turning radius input to ensure comfortable turns from row to row. The working area is finally divided into rows with the adequate working width and eventual overlap. In Fig. 1, a schematic representation of the processing procedure of the ORP tool is depicted. A common practise is to define the rows across the working area as being parallel to the longest edge of the field; however in some cases this may provide a suboptimal solution (Hameed, Bocthis, & Sorensen, 2011). Furthermore, since one of the goals of the tool is to minimise the traffic in the working area, all traversing between rows is limited to the headlands, i.e. all rows must terminate on a headland. Theoretically, there are an infinite number of candidate working directions for the rows, however to ease processing the proposed candidates are limited based on the field boundaries. After a set of rows has been created for each candidate driving direction,
151
each set is evaluated by an optimisation algorithm, based on number rows and overlap, in order to find the optimised driving direction. The working rows are plotted as rectangles across the working area (see Fig. 2). Once the optimal driving direction is found with respect to model restrictions, the optimised route required to cover all working rows and headlands is created based on a combinatorial optimisation algorithm. The connection paths are the paths followed by the machine to connect the working rows. Connection paths are defined linking: each extremity of each working row to any other extremity of a working row; each gate and to each headland; each headland to each other headland; each headland to each working row extremity; and each working row extremity to each gate. In addition, the connection paths are constrained by the minimum turning radius of the machine so that they are traversable by the vehicle. Once the connection paths are defined, an optimisation algorithm finds the optimised route to minimise the distance of the connection paths so that all the working rows and headlands are travelled on, i.e. all the field is mowed. Contrary to the working paths, all the connection paths do not necessary need to be followed to complete the work. The algorithms attempt to find the shortest possible non-working distance for the machine, however due to the complexity and size of the optimisation problem, the final optimised route can only be described as near optimal.
2.2.
Experimental conditions
The field recording took place in the surroundings of Lem, in western part of Jutland in Denmark, where between the 12th and 14th of October approximately 120 ha were mowed by an experienced and highly skilled machine operator. The fields were located between 56º000 3600 N and 56º040 4800 N, and between 8º180 0000 E and 8º270 3600 E. The mowing operation was performed using a CLAAS Xerion 5000 tractor with a front mounted CLAAS 9300 Duo Cutter with an optimal working
Fig. 1 e Diagram showing the processing procedure followed by the ORP tool.
152
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Fig. 2 e Graphical representation of two headlands (in yellow), working area (in blue) and the working rows plotted as rectangles (in green). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
width of 6.64 m (see Fig. 3). So as not to influence the results, the operator was only instructed where to place GPS recorder and was asked to execute the operation as they normally would. The records were collected by a Qstarz Travel Recorder XT (Taipei, Taiwan), which is a high sensitive (approx. 165 dB) GPS recorder, in order to record the position data. The GPS recorder is also able to receive an A-GPS signal providing improving the <2.5 m accuracy to 30e40 cm. The GPS recorders were mounted inside the cabin to the steering console, ensuring good signal reception. Latitude, longitude and time stamp were recorded in standard National Marine Electronics Association (NMEA) text strings with a frequency of 1 Hz and afterwards transferred to the computer in form of txt files.
3.
Evaluation analysis procedure
The latitude and longitude data were converted to the Universal Transverse Mercator (UTM) so that the data could be plotted in metres on a flat surface.
As the GPS recorder collected data continuously for several days, the data was divided into datasets according to each field by visual inspection. The stationary points that were recorded were caused by the operator setting up the machine or taking a break, and were therefore manually discarded from the datasets. Two of the required inputs for the ORP tool to create the optimised route are the Fb and the Fg. Overlaying a map and manually tracing the boundaries and gates can be cause for biases and errors; therefore the headlands recorded by the GPS were found manually by visual inspection of the data points, so that the field boundary could be calculated and inversely the headlands could now be generated by the ORP tool. The Fg were determined from the first and last points recorded by the machine within the field boundary and then transposed onto the nearest boundary, as all Fg must be on the boundary. When the gates were closer than 5 m from each other, they were considered as one single gate. Once the field boundaries and gates were determined, the working width and turning radius were introduced, so that the ORP tool was able to find the optimised route plans for all the fields. It should be noted that on some fields the working width was adjusted, so that the number of rows of the optimised route were in line with number recorded by the GPS, in order to minimise biases from overlap and thus improve the comparison of the two. For example, the measured working widths in the field varied from 5.81 to 8.72 m, compared to the optimal working width of the machine which is 6.64 m. The measured wider working width than stipulated could be caused by non-notified changes in the machinery or by purposely left uncut strips in the field; nonetheless, it required adjustments in the ORP tool so that the comparison was reliable. To evaluate the ORP tool, distance was used to compare the data recorded by the GPS and the data generated by the tool, as distance is proportional to the variable cost of operation, e.g. fuel or labour costs. Time could have been used as well, however the removed data points from the GPS recorder made time inconsistent and the operational speed of the machine was not uniform, making it unviable to compare the data using this parameter. In addition to the comparison of the recorded and optimised distance, the traffic intensity on the field was evaluated. The traffic intensity has been measured as the number of passes by the machine over a field area of size 5 5 m, which is slightly smaller than the minimum working width (5.81 m) to avoid false readings. For this evaluation, the field has been divided into headlands and working area in order to observe differences in the trafficking intensity on each part of the field.
4.
Fig. 3 e Mowing operations recorded.
Results
The GPS recorder collected data continuously; therefore, the fields were divided into datasets and labelled by visual inspection as depicted in Figs. 4 and 5. Twelve fields were successfully recorded. Seven fields were not included in the evaluation as five fields had large wet areas in which the tractor could not perform any operation, and two other fields had areas deemed too narrow (less than twice the working
153
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Table 1 e Recorded and optimised distance comparison. Driven Optimised Difference Field (%) area (m2) distance (m) distance (m) Field Field Field Field Field Field Field Field Field Field Field Field
Fig. 4 e Continuous GPS recordings and field labelling of first set of recordings.
width across), which would hinder how the ORP tool created the headlands. Table 1 shows the distance recorded by the GPS during the field operations and the optimised distance calculated by the ORP tool. The table shows as well the excess distance in percentage between the operator's route and the route proposed by the ORP tool. And the area of each labelled field is also included. The operator travelled between 0.2% and 18.4% more distance in each field than the optimised distance calculated by the ORP tool, with an average increase of 8.92%. In total from the twelve fields studied, the ORP tool saved 9.2 km of driving distance of a total of 123 km driven by the operator, i.e. a saving of 7.5%. The route recorded by the machine and the route designed by the ORP tool on field 7 are plotted in Fig. 6. Field 7 has been chosen as an example field, because it is one of the largest fields, and has a total distance difference of 9.3%, which is close to the average distance savings.
Fig. 5 e Continuous GPS recordings and field labelling of second set of recordings.
1 2 3 4 5 6 7 8 9 10 11 12
49,913.42 38,438.32 8838.83 12,230.06 57,057.32 37,247.75 111,498.45 27,956.51 124,360.62 100,486.66 31,674.64 74,576.51
9713.80 8044.20 1850.10 2438.20 11,959.00 8323.40 24,024.00 5202.70 18,259.00 15,811.00 5939.80 11,165.00
9345.52 8005.96 1563.73 2433.87 10,137.22 7773.31 21,984.16 4393.88 17,452.77 14,891.23 5153.83 10,401.60
3.9 0.5 18.3 0.2 18.0 7.1 9.3 18.4 5.6 6.2 15.3 7.3
Histograms showing the number of passes per 5 5 m area in field 7 are depicted in Figs. 7 and 8. The traffic intensity of both the operator's route and the ORP tool during the field operation has been divided into headlands and working area. Table 2 shows data for the trafficking intensity for all twelve fields, distinguishing headlands and working area. The data shows mean and variance values for the number of passes per area in each field. All the fields have a lower mean number of passes per area in both headlands and working area, except for a few were it is equal, with a 95% of confidence interval, for the ORP route compared to the operator's route. The mean values and variances for the whole field were also lower for the optimised route, being up to 0.14 passes per area lower, with the exception of two of the smallest fields where it was equal, with a 95% of confidence interval.
5.
Discussion
The ORP tool reduced the infield travelled distance in all tested fields (see Table 1), with the routes managed by the operator being up to 18.4% longer, saving consequently fuel and hypothetically time during the field operations. Time savings have to be considered hypothetical as the evaluation did not consider speed or acceleration, which might have affected the total time savings of the operation. Including speed in the comparison was not possible, as stated before, because of non-uniform operational speed and breaks taken in the field. Time savings are especially considerable during the connection paths in the headland, as the smoothness of the turn affects directly the speed. Therefore, in order to save time, the optimising algorithm of the ORP tool guides the tractor not to reverse, in order to keep the momentum and avoid extra wheel stress exertion on the soil. However, some studies analysing smooth turning paths in headlands, where speed and acceleration were taken into consideration, showed that reversing was faster than forward turning for distances below 10 m in the x direction (Backman, Piirainen, & Oksanen, 2015; € ben, Meyer zu Helligen, & Schulze Lammers, Sabelhaus, Ro 2013). Even though this size of turning distance, i.e. <10 m, is sometimes modelled by the ORP tool in the optimised routes, the time saved is just of a few seconds per turn, which
154
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Fig. 6 e Plotted operating routes on Field 7 by GPS recordings (left) and by the ORP tool (right).
does not indicate considerable overall time savings. Moreover, not reversing makes the driving easier for the operator. It can therefore be assumed that the ORP tool reduces not only the travelled distance, but also the overall operational time. The operational savings a farmer or agricultural contractor could achieve are large by implementing optimised routes, when it is considered that in a silage growing season a given field may be mowed up to 5 times. Moreover, other operations such as raking and baling could also take advantage of the ORP to find optimised routes. It should also be noted that the comparison is made between the ORP tool and a highly skilled driver, suggesting that if a less skilled driver had been used, the savings gained by using the ORP tool would have been greater. This could create an opportunity for the ORP tool to be used as a training tool for inexperienced drivers, or for a farmer or agricultural contractor to employ a cheaper work force.
The business model employed by agricultural contractors could offer them a unique possibility to gain an advantage using the ORP tool. Currently, farmers pay agricultural contractor to carry out operations on a per area basis, while an agricultural contractor will pay their workers on an hourly basis (Edwards, 2015). If the ORP can be used to reduce the time needed to process an area then the agricultural contractors costs will decrease while their income will remain the same, leading to increased profits. Traffic intensity was also evaluated looking separately into headlands and working areas with the intention of observing how the traffic intensity is distributed in the field (see Table 2). In most fields, the routes proposed by the ORP tool significantly reduced the traffic intensity in the working area, except for three fields, where the number of passes per area was equal, with a 95% of confidence interval. These three fields where no reduction of traffic intensity was made, are some of
Fig. 7 e Histogram showing the number of passes per area in the working area of Field 7.
Fig. 8 e Histogram showing the number of passes per area in the headlands of Field 7.
155
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Table 2 e Recorded and optimised traffic intensity comparison. Working area
Field Field Field Field Field Field Field Field Field Field Field Field
1 2 3 4 5 6 7 8 9 10 11 12
Headlands
GPS mean
GPS var
ORP mean
ORP var
GPS mean
GPS var
ORP mean
ORP var
0.8422 0.8194 0.6484 0.5029 0.8858 0.7912 0.8685 0.4925 0.5184 0.5118 0.4822 0.5646
0.3394 0.3964 0.4502 0.2514 0.3874 0.2973 0.3510 0.2542 0.2699 0.2739 0.3095 0.3322
0.6174 0.7320 0.5641 0.5205 0.8120 0.7574 0.8115 0.5171 0.4936 0.4621 0.5015 0.5256
0.2438 0.2431 0.2575 0.2746 0.1797 0.2060 0.1700 0.2574 0.2500 0.2494 0.2504 0.2495
0.7553 0.7463 0.8455 0.9204 0.8878 1.0606 1.0815 0.9586 0.7206 0.8131 0.9660 0.7783
0.6036 0.8282 1.0672 1.0386 0.9216 1.3674 1.5007 1.1356 0.6025 0.7677 0.9424 0.6135
0.7119 0.8020 0.8351 0.8945 0.8691 1.0927 1.1221 0.8228 0.7281 0.8114 0.9408 0.7633
0.5086 0.6941 0.8417 0.8176 0.8309 1.1618 1.3984 0.8377 0.7266 0.9735 1.1900 0.7264
the smallest fields recorded, which may suggest that the operator and the ORP tool performed equally well, regarding traffic intensity in the working area. In addition, even if the ORP tool confines all the non working distance to the headlands, most of the fields still showed a reduction of mean traffic intensity in the headlands. This displacement of traffic from working area to headlands can be observed for field 7 in the histograms shown in Figs. 7 and 8. It can be seen that the distribution of the number of passes per area is relocated to lower values in the working area, as it assigns all the nonworking distance to the headlands. Consequently the distribution of the number of passes is slightly relocated to higher values in the headlands. In general terms, the routes proposed by the ORP tool for the tested fields had lower means and variances of number of passes per area, meaning lower traffic intensity. An aspect that the traffic intensity evaluation did not consider is the actual wheel traffic. Traffic intensity calculations were made using the GPS position during the operation, so the number of passes per area reflects the area covered by the machine, rather than the compacted area by wheel pressure. It can therefore be observed in the histograms in Figs. 7 and 8, that there are a considerable number of areas with no GPS readings indicating no passes. In general terms, the routes proposed by the ORP tool had lower means and variances in the number of passes per area, meaning lower traffic intensity. This is especially true for the working area, where considerable traffic intensity is reduced, as the ORP tool allocates all non-working distance to the headlands (see Fig. 6). The ORP tool prototype calculates the optimised route considering the headlands along the whole field's borders and not only the top and bottom headlands. This is an interesting feature that many route planning optimisations do not consider (Bakhtiari, Navid, Mehri, & Bochtis, 2011; Bochtis & ~ oz, Vougioukas, 2008; Bochtis et al., 2010; Conesa-Mun Pajares, & Ribeiro, 2016; Jensen et al., 2012). This feature enhances the applicability of optimisation algorithms from theory to practice. The selection of driving direction and division of working area into rows can be a difficult task for operators, especially in large fields. The optimisation algorithm of the ORP tool provides enhanced solutions for the driving direction and row
division, as it is clearly depicted in Fig. 6, where the number of turns inside the working area was visibly reduced. There are however some limitations to the ORP tool. The precision of the tool is dependent on the precision and accuracy of the satellite navigation system. For example, the precision can vary from a few metres, in a cheap GPS receiver, to 1 cm in Real Time Kinematic (RTK) navigation. While this accuracy is independent of the actual ORP tool, it may affect the optimal route calculation. Also, the ORP tool prototype has some existing limitations that need development. Although the tool is designed for NMF operations, the current tool would not be able to optimise ploughing operations, which require a special route pattern in order to fill the empty furrow to the left, by a new furrow during the next pass. Ploughing is the operation that requires highest energy inputs and consequently emits more greenhouse gasses (Dyer & , Petrovic , & Ðevic , 2010), and the Desjardins, 2003; Mileusnic process that causes important subsoil compaction problems (Alakukku et al., 2003). Therefore, this type of operation should be included in the design to fully exploit the possibilities of the tool, and increase sustainability and profitability for the farmer. Fields with narrow areas can present complications for finding an optimised route using the ORP tool. This limitation did not allow finding an optimised route for two of the fields tested which had areas narrower than twice the working width of the mower. Finally, fields with obstacles or considerable slopes will present substantial challenges for the driver to follow the optimised route, which currently limits the applicability of the tool to relatively flat fields with absence or small sized obstacles. Since the ORP tool helps reducing the infield travelled distance and trafficking intensity, it can also be used as a tool for reducing soil compaction. Nevertheless, soil compaction has to be prevented by a combination of different approaches, e.g. increasing soil organic matter (De Neve & Hofman, 2000), adequate crop rotation (Hamza & Anderson, 2005), reducing pressure on soil by low tyre inflation (Alakukku et al., 2003), or working in the field under soil ready conditions (Edwards, , 2016), which can White, Munkholm, Sørensen, & Lamande be combined with the ORP tool. Additional information, such as soil types/moisture, weather, vehicle axle loading, etc., would also need to be made available to the tool so that it could include the consequences of trafficking different areas
156
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
of the field at different times within its optimisation algorithms. The feature of the tool to confine all non-working activity to the headlands makes the prototype a potential tool for controlled traffic farming, as it reduces the traffic in the working area of a field to a minimum. Furthermore, optimised route planners such as the ORP tool, will become indispensable for precision farming in the near future, since they can reduce production costs and operational time, making farming operations more efficient. In general terms, the tool can benefit multiple types of farmers in making their NMF operations more effective, no matter if they practise conventional, organic or precision farming, by reducing the travelled distance in the field by up to 18%. A final advantage of the ORP tool is its development as an application for mobile operating systems, which makes it easy for many farmers to acquire and use it.
6.
Conclusion
The evaluation of the ORP tool prototype was performed comparing travelled distances and traffic intensity by a professional operator during mowing operations and by the optimised route proposed by the tool. In all twelve fields compared, the operator travelled longer distances than the ORP tool's proposed distance, being up to 18.4% longer, with a reduction of 9.2 km (7.5%) in the total infield travelled distance. The tool also generally reduced the traffic intensity, especially in the working area, as it confines all non-working distance to the headlands. Further work is still needed for the ORP tool to be able to fully optimise additional operations, such as ploughing which is the operation with the highest energy requirements and one that significantly affects subsoil compaction. The tool does also present complications when optimising routes for fields with very narrow areas.
references
Alakukku, L. (1996). Persistence of soil compaction due to high axle load traffic. II. Long- term effects on the properties of finetextured and organic soils. Soil and Tillage Research, 37(4), 223e238. http://dx.doi.org/10.1016/0167-1987(96)01017-3. Alakukku, L., Weisskopf, P., Chamen, W. C. T., Tijink, F. G. J., Van Der Linden, J. P., Pires, S., et al. (2003). Prevention strategies for field traffic-induced subsoil compaction: A review part 1. Machine/soil interactions. Soil and Tillage Research, 73(1e2), 145e160. http://dx.doi.org/10.1016/S0167-1987(03)00107-7. Backman, J., Piirainen, P., & Oksanen, T. (2015). Smooth turning path generation for agricultural vehicles in headlands. Biosystems Engineering, 139, 76e86. http://dx.doi.org/10.1016/ j.biosystemseng.2015.08.005. Bakhtiari, A., Navid, H., Mehri, J., & Bochtis, D. D. (2011). Optimal route planning of agricultural field operations using ant colony optimization. Agricultural Engineering International: CIGR Journal, 13(4), 1e16. Retrieved from http://www.cigrjournal. org/index.php/Ejounral/article/view/1939. Bhandral, R., Saggar, S., Bolan, N. S., & Hedley, M. J. (2007). Transformation of nitrogen and nitrous oxide emission from
grassland soils as affected by compaction. Soil and Tillage Research, 94(2), 482e492. http://dx.doi.org/10.1016/ j.still.2006.10.006. Bochtis, D. D., & Sørensen, C. G. (2009). The vehicle routing problem in field logistics part I. Biosystems Engineering, 104(4), 447e457. http://dx.doi.org/10.1016/ j.biosystemseng.2009.09.003. Bochtis, D. D., Sørensen, C. G., Busato, P., & Berruto, R. (2013). Benefits from optimal route planning based on B-patterns. Biosystems Engineering, 115(4), 389e395. http://dx.doi.org/ 10.1016/j.biosystemseng.2013.04.006. Bochtis, D. D., Sørensen, C. G., & Vougioukas, S. G. (2010). Path planning for in-field navigation-aiding of service units. Computers and Electronics in Agriculture, 74(1), 80e90. http:// dx.doi.org/10.1016/j.compag.2010.06.008. Bochtis, D. D., & Vougioukas, S. G. (2008). Minimising the nonworking distance travelled by machines operating in a headland field pattern. Biosystems Engineering, 101(1), 1e12. http://dx.doi.org/10.1016/j.biosystemseng.2008.06.008. ~ oz, J., Pajares, G., & Ribeiro, A. (2016). Mix-opt: A new Conesa-Mun route operator for optimal coverage path planning for a fleet in an agricultural environment. Expert Systems with Applications, 54, 364e378. http://dx.doi.org/10.1016/j.eswa.2015.12.047. De Neve, S., & Hofman, G. (2000). Influence of soil compaction on carbon and nitrogen mineralization of soil organic matter and crop residues. Biology and Fertility of Soils, 30(5e6), 544e549. Dyer, J. A., & Desjardins, R. L. (2003). Simulated farm fieldwork, energy consumption and related greenhouse gas emissions in Canada. Biosystems Engineering, 85(4), 503e513. http:// dx.doi.org/10.1016/S1537-5110(03)00072-2. Edwards, G. T. C. (2015). Field readiness and operation scheduling. Mechanical Engineering. Aarhus University. Edwards, G., White, D. R., Munkholm, L. J., Sørensen, C. G., & , M. (2016). Modelling the readiness of soil for Lamande different methods of tillage. Soil and Tillage Research, 155, 339e350. http://dx.doi.org/10.1016/j.still.2015.08.013. EU. (2016). Short-term outlook for EU arable crops, dairy and meat markets in 2016 and 2017. Retrieved from http://ec.europa.eu/ agriculture/markets-and-prices/short-term-outlook/pdf/20163_en.pdf. FAO. (2015). The state of agricultural commodity markets 2015e16. Trade and food security: Achieving a better balance between national priorities and the collective good (3rd ed.). Rome: Food and Agriculture Organization of the United Nations. Retrieved from http://www.fao.org/3/a-i5090e.pdf. Gła˛b, T. (2014). Effect of soil compaction and N fertilization on soil pore characteristics and physical quality of sandy loam soil under red clover/grass sward. Soil and Tillage Research, 144, 8e19. Hameed, I. A., Bocthis, D. D., & Sorensen, C. G. (2011). Driving angle and track sequence optimization for operational path planning using genetic algorithms. Applied Engineering in Agriculture, 26(6), 1077e1086. http://dx.doi.org/10.13031/ 2013.40615. Hamza, M. A., & Anderson, W. K. (2005). Soil compaction in cropping systems: A review of the nature, causes and possible solutions. Soil and Tillage Research, 82(2), 121e145. http:// dx.doi.org/10.1016/j.still.2004.08.009. ska-Jurkiewicz, A., & van _ Horn, R., Domz_ zał, H., Słowin Ouwerkerk, C. (1995). Soil compaction processes and their effects on the structure of arable soils and the environment. Soil and Tillage Research, 35(1e2), 23e36. http://dx.doi.org/ 10.1016/0167-1987(95)00479-C. Jensen, M. A. F., Bochtis, D., Sorensen, C. G., Blas, M. R., & Lykkegaard, K. L. (2012). In-field and inter-field path planning for agricultural transport units. Computers and Industrial Engineering, 63(4), 1054e1061. http://dx.doi.org/10.1016/ j.cie.2012.07.004.
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Kaivosoja, J., Jackenkroll, M., Linkolehto, R., Weis, M., & Gerhards, R. (2014). Automatic control of farming operations based on spatial web services. Computers and Electronics in Agriculture, 100, 110e115. http://dx.doi.org/10.1016/ j.compag.2013.11.003. Keller, T., Arvidsson, J., Dawidowski, J. B., & Koolen, A. J. (2004). Soil precompression stress: II. A comparison of different compaction tests and stress-displacement behaviour of the soil during wheeling. Soil and Tillage Research, 77(1), 97e108. http://dx.doi.org/10.1016/j.still.2003.11.003. la, F., H Kroulı´k, M., Kumha ula, J., & Honzı´k, I. (2009). The evaluation of agricultural machines field trafficking intensity for different soil tillage technologies. Soil and Tillage Research, 105(1), 171e175. http://dx.doi.org/10.1016/j.still.2009.07.004. Lewis, T. (1998). Evolution of farm management information systems. Computers and Electronics in Agriculture, 19(3), 233e248. http://dx.doi.org/10.1016/S0168-1699(97)00040-9. McHugh, A. D., Tullberg, J. N., & Freebairn, D. M. (2009). Controlled traffic farming restores soil structure. Soil and Tillage Research, 104(1), 164e172. http://dx.doi.org/10.1016/j.still.2008.10.010.
157
, Z. I., Petrovic , D. V., & Ðevic , M. S. (2010). Comparison Mileusnic of tillage systems according to fuel consumption. Energy, 35(1), 221e228. http://dx.doi.org/10.1016/j.energy.2009.09.012. € ben, F., Meyer zu Helligen, L. P., & Schulze Sabelhaus, D., Ro Lammers, P. (2013). Using continuous-curvature paths to generate feasible headland turn manoeuvres. Biosystems Engineering, 116(4), 399e409. http://dx.doi.org/10.1016/ j.biosystemseng.2013.08.012. Sørensen, C. G., & Bochtis, D. D. (2010). Conceptual model of fleet management in agriculture. Biosystems Engineering, 105(1), 41e50. http://dx.doi.org/10.1016/j.biosystemseng.2009.09.009. Tilman, D., Cassman, K. G., Matson, P. A., Naylor, R., & Polasky, S. (2002). Agricultural sustainability and intensive production practices. Nature, 418(6898), 671e677. http://dx.doi.org/ 10.1038/nature01014. Uchida, Y., Clough, T. J., Kelliher, F. M., & Sherlock, R. R. (2008). Effects of aggregate size, soil compaction, and bovine urine on N2O emissions from a pasture soil. Soil Biology and Biochemistry, 40(4), 924e931. http://dx.doi.org/10.1016/ j.soilbio.2007.11.007.
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/issn/15375110
Research Paper
Route planning evaluation of a prototype optimised infield route planner for neutral material flow agricultural operations Gareth T.C. Edwards a,*, Jørgen Hinge b, Nick Skou-Nielsen a, Andres Villa-Henriksen a, Claus Aage Grøn Sørensen c, Ole Green a a
Agro Intelligence, Agro Food Park 13, 8200 Aarhus N, Denmark AgroTech A/S, Agro Food Park 15, 8200 Aarhus N, Denmark c Aarhus University, Nordre Ringgade 1, 8000 Aarhus C, Denmark b
article info
The need to decrease unit production costs has led agricultural industries to develop larger
Article history:
and consequently heavier machinery. While this has increased the productivity of single
Received 1 July 2016
machines, it has also caused significant soil compaction, which may cause reduced crop yield
Received in revised form
and crop quality. Therefore, mechanisation solutions that have both lower unit costs and
7 September 2016
reduce the risk of soil compaction are needed. Optimising infield routes will reduce labour
Accepted 11 October 2016
costs, fuel consumption and field trafficking intensity, providing important benefits for
Published online 11 December 2016
infield operations. In this paper, a prototype of an optimised infield route planning tool for neutral material flow operations is evaluated. The evaluation parameters focused on dis-
Keywords:
tance and traffic intensity reductions, comparing the routes proposed by the tool prototype
Optimised infield route planner
and the routes followed by a professional operator during mowing operations. The tool re-
Path planning
quires some minimum inputs: field boundaries, field gates, working width and minimum
Performance evaluation
turning radius, in order to provide an optimised route. Twelve fields were recorded by a Global
Travelled distance reduction
Positioning System (GPS) during mowing operations and later compared with the routes
Trafficking intensity
proposed by the tool. In all fields, the operator's normal route was longer in distance than the
Precision agriculture
route proposed by the tool, being up to 18.4% longer. In total, 9.2 km of infield distance was saved, i.e. 7.5%. The traffic intensity was reduced in all fields, except for two of the smallest fields, where it equalled that of the normal route. Specifically, the traffic intensity was reduced in the working areas, as the tool confined all non-working distance to the headlands. © 2016 The Authors. Published by Elsevier Ltd on behalf of IAgrE. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.
Introduction
The increasing world demand on agricultural products has led the farming industry to increase productivity by diverse
solutions from different disciplines, e.g. genetics, agronomy or engineering (Tilman, Cassman, Matson, Naylor, & Polasky, 2002). In the last decades, the engineering focus has been to develop large, powerful, and high-capacity machinery, in order to decrease unit costs. However, this development has
* Corresponding author. E-mail address: [email protected] (G.T.C. Edwards). http://dx.doi.org/10.1016/j.biosystemseng.2016.10.007 1537-5110/© 2016 The Authors. Published by Elsevier Ltd on behalf of IAgrE. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
150
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Nomenclature GPS ORP NMF Ww Rm Fg Fb Ph Pr Pc
Global Positioning System Optimised Infield Route Planner Neutral Material Flow Working width Minimum turning radius Field gates position Field boundaries position Headland paths Row paths Connection paths
made machinery heavier, and it is consequently creating important subsoil compaction problems, which may result in lower yields and reduced fertiliser efficiency, along with higher water-logging, run-off and erosion problems (Hamza & Anderson, 2005). Soil degradation is an increasing problem worldwide, mainly caused in modern agriculture by subsoil compaction caused by heavy machinery (Hamza & Anderson, 2005). Ac la, H cording to Kroulı´k, Kumha ula, and Honzı´k (2009), p. 95% of the area in a field in conventional agriculture is run over at least once a year, meaning that the problem is generalised throughout the field. Repeated traffic over an area highly increases the risk (Keller, Arvidsson, Dawidowski, & Koolen, 2004). Soil compaction means increased bulk density and homogenisation of the soil, causing decreased aeration and water infiltration and increased penetration resistance, which impedes a proper root development and limits the biological ska_ activity of the soil (Gła˛b, 2014; Horn, Domz_ zał, Słowin Jurkiewicz, & van Ouwerkerk, 1995). Subsoil compaction has long term effects on the soil which are difficult to solve, therefore the most effective practice is to avoid or reduce the compaction as much as possible, rather than apply post-effect solutions, e.g. deep ripping (Laura Alakukku, 1996). Furthermore, soil compaction also has effects on greenhouse gas emissions. For example, Bhandral, Saggar, Bolan, and Hedley (2007) found in grassland soils that the N2O emissions were between 2 and 14 times higher for a compacted soil than for an uncompacted soil, with the rate for nitrate fertilisation being especially high. These results were corroborated by Uchida, Clough, Kelliher, and Sherlock (2008). Regardless of compaction problems, farmers need to decrease unit costs in order to adapt to and compete in the globalised modern market system, where apart from the exceptional increases in 2007e08, and 2011, low prices dictate their agenda (EU, 2016; FAO, 2015). There is therefore a need for solutions that can reduce both the unit costs and reduce the risk of soil compaction. Computer based tools can both optimise farming operations, as well as minimise the risks from soil compaction, making the whole system more sustainable. Although computer innovation in farming has been more common for business related activities than specifically for farming (Lewis, 1998), in the last decade the number of computer based tools in precision farming has grown considerably (Kaivosoja, Jackenkroll, Linkolehto, Weis, & Gerhards, 2014). One of these tools is an infield route planner, which optimises the route followed by the
machinery. Optimised infield route planning can reduce labour costs and fuel consumption, in addition to reduce field trafficking, which is one of several solutions proposed to reduce soil compaction (Bochtis, Sørensen, Busato, & Berruto, 2013; Hamza & Anderson, 2005). Moreover, infield route planning can be used for controlled traffic farming, which offers farmers an opportunity to not only reduce soil compaction, but also to restore their soils (McHugh, Tullberg, & Freebairn, 2009). Optimised route planners are already commonly used in Global Positioning System (GPS) based applications in mobiles and computer systems, for personal use, as well as for industrial and logistical uses; however they are still in the development stages for farming applications (Sørensen & Bochtis, 2010). Several studies and projects are working with optimised algorithms and solutions for route planning in precision agriculture, and especially for the emerging development of field robots. Different activities require different approaches, as some activities have capacity constraints and may require aid from support units (Bochtis, Sørensen, & Vougioukas, 2010), e.g. cistern trucks for replenishing fertilisers or pesticides. Furthermore, there may be in-field and inter-field routes that need to be optimised, making route planning in agriculture a challenging task, as there are infield attributes (e.g. working tracks and headland passes), and inter-field configurations (e.g. field gates and road networks) to be considered (Jensen, Bochtis, Sorensen, Blas, & Lykkegaard, 2012). However, none of the above mentioned models have been adapted into functional tools which can be used by a tractor drive to optimise in field operations in realtime. An optimised Infield Route Planner (ORP) prototype tool is evaluated. The ORP tool is designed for ‘neutral material flow’ (NMF) field operations, i.e. operations where there is no flow of material into or out of the field (e.g. tillage, cultivation, mowing) (Bochtis & Sørensen, 2009). Here, the tool is evaluated by comparing the distance travelled by a professional operator during mowing operations, with the optimised distance proposed by the ORP tool. Optimised route planners can save travelling distance inherent in agricultural operations, reducing consequently unit costs, and compaction risks. Secondarily, the ORP tool is evaluated by comparing the trafficking intensity as derived from the recorded data and the modelled optimised route, in order to assess the potential efficacy of the tool for reducing soil compaction.
2.
Material and methods
Applied mathematical modelling provides many possibilities to improve the effectiveness of precision agriculture. In this case, the driving pattern of agricultural machinery in the field is optimised by the application of an ORP prototype tool. The developed ORP tool has been evaluated by comparing the routes followed by an experienced tractor driver during mowing operations and the ones proposed by the ORP. The tool and the conditions are explained below.
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
2.1.
Description of the ORP tool
The Optimised Infield Route Planner (ORP) tool is designed as an application software and was developed by Agro Intelligence Aps, Aarhus, Denmark. The application creates routes for machines to follow when executing infield operations and with the aim of minimising the distance travelled by the machine. For this, the developed application requires the following inputs: field boundaries position (Fb), field gates position (Fg), working width (including any possible overlap) (Ww), and minimum turning radius of the machine (Rm). Once the inputs are given, the field is automatically divided into headlands and working rows, with their corresponding headland paths (Ph), row paths (Pr) and connection paths (Pc). The field boundaries are automatically registered as the outer edge of the working width during the operator's driving through the first track in the headland. The gates are also registered as the entry point of the machine in the field. The inner edge of the first headland defines the second and subsequent headlands, as well as the main working area. The number of headlands is calculated according to the turning radius input to ensure comfortable turns from row to row. The working area is finally divided into rows with the adequate working width and eventual overlap. In Fig. 1, a schematic representation of the processing procedure of the ORP tool is depicted. A common practise is to define the rows across the working area as being parallel to the longest edge of the field; however in some cases this may provide a suboptimal solution (Hameed, Bocthis, & Sorensen, 2011). Furthermore, since one of the goals of the tool is to minimise the traffic in the working area, all traversing between rows is limited to the headlands, i.e. all rows must terminate on a headland. Theoretically, there are an infinite number of candidate working directions for the rows, however to ease processing the proposed candidates are limited based on the field boundaries. After a set of rows has been created for each candidate driving direction,
151
each set is evaluated by an optimisation algorithm, based on number rows and overlap, in order to find the optimised driving direction. The working rows are plotted as rectangles across the working area (see Fig. 2). Once the optimal driving direction is found with respect to model restrictions, the optimised route required to cover all working rows and headlands is created based on a combinatorial optimisation algorithm. The connection paths are the paths followed by the machine to connect the working rows. Connection paths are defined linking: each extremity of each working row to any other extremity of a working row; each gate and to each headland; each headland to each other headland; each headland to each working row extremity; and each working row extremity to each gate. In addition, the connection paths are constrained by the minimum turning radius of the machine so that they are traversable by the vehicle. Once the connection paths are defined, an optimisation algorithm finds the optimised route to minimise the distance of the connection paths so that all the working rows and headlands are travelled on, i.e. all the field is mowed. Contrary to the working paths, all the connection paths do not necessary need to be followed to complete the work. The algorithms attempt to find the shortest possible non-working distance for the machine, however due to the complexity and size of the optimisation problem, the final optimised route can only be described as near optimal.
2.2.
Experimental conditions
The field recording took place in the surroundings of Lem, in western part of Jutland in Denmark, where between the 12th and 14th of October approximately 120 ha were mowed by an experienced and highly skilled machine operator. The fields were located between 56º000 3600 N and 56º040 4800 N, and between 8º180 0000 E and 8º270 3600 E. The mowing operation was performed using a CLAAS Xerion 5000 tractor with a front mounted CLAAS 9300 Duo Cutter with an optimal working
Fig. 1 e Diagram showing the processing procedure followed by the ORP tool.
152
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Fig. 2 e Graphical representation of two headlands (in yellow), working area (in blue) and the working rows plotted as rectangles (in green). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
width of 6.64 m (see Fig. 3). So as not to influence the results, the operator was only instructed where to place GPS recorder and was asked to execute the operation as they normally would. The records were collected by a Qstarz Travel Recorder XT (Taipei, Taiwan), which is a high sensitive (approx. 165 dB) GPS recorder, in order to record the position data. The GPS recorder is also able to receive an A-GPS signal providing improving the <2.5 m accuracy to 30e40 cm. The GPS recorders were mounted inside the cabin to the steering console, ensuring good signal reception. Latitude, longitude and time stamp were recorded in standard National Marine Electronics Association (NMEA) text strings with a frequency of 1 Hz and afterwards transferred to the computer in form of txt files.
3.
Evaluation analysis procedure
The latitude and longitude data were converted to the Universal Transverse Mercator (UTM) so that the data could be plotted in metres on a flat surface.
As the GPS recorder collected data continuously for several days, the data was divided into datasets according to each field by visual inspection. The stationary points that were recorded were caused by the operator setting up the machine or taking a break, and were therefore manually discarded from the datasets. Two of the required inputs for the ORP tool to create the optimised route are the Fb and the Fg. Overlaying a map and manually tracing the boundaries and gates can be cause for biases and errors; therefore the headlands recorded by the GPS were found manually by visual inspection of the data points, so that the field boundary could be calculated and inversely the headlands could now be generated by the ORP tool. The Fg were determined from the first and last points recorded by the machine within the field boundary and then transposed onto the nearest boundary, as all Fg must be on the boundary. When the gates were closer than 5 m from each other, they were considered as one single gate. Once the field boundaries and gates were determined, the working width and turning radius were introduced, so that the ORP tool was able to find the optimised route plans for all the fields. It should be noted that on some fields the working width was adjusted, so that the number of rows of the optimised route were in line with number recorded by the GPS, in order to minimise biases from overlap and thus improve the comparison of the two. For example, the measured working widths in the field varied from 5.81 to 8.72 m, compared to the optimal working width of the machine which is 6.64 m. The measured wider working width than stipulated could be caused by non-notified changes in the machinery or by purposely left uncut strips in the field; nonetheless, it required adjustments in the ORP tool so that the comparison was reliable. To evaluate the ORP tool, distance was used to compare the data recorded by the GPS and the data generated by the tool, as distance is proportional to the variable cost of operation, e.g. fuel or labour costs. Time could have been used as well, however the removed data points from the GPS recorder made time inconsistent and the operational speed of the machine was not uniform, making it unviable to compare the data using this parameter. In addition to the comparison of the recorded and optimised distance, the traffic intensity on the field was evaluated. The traffic intensity has been measured as the number of passes by the machine over a field area of size 5 5 m, which is slightly smaller than the minimum working width (5.81 m) to avoid false readings. For this evaluation, the field has been divided into headlands and working area in order to observe differences in the trafficking intensity on each part of the field.
4.
Fig. 3 e Mowing operations recorded.
Results
The GPS recorder collected data continuously; therefore, the fields were divided into datasets and labelled by visual inspection as depicted in Figs. 4 and 5. Twelve fields were successfully recorded. Seven fields were not included in the evaluation as five fields had large wet areas in which the tractor could not perform any operation, and two other fields had areas deemed too narrow (less than twice the working
153
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Table 1 e Recorded and optimised distance comparison. Driven Optimised Difference Field (%) area (m2) distance (m) distance (m) Field Field Field Field Field Field Field Field Field Field Field Field
Fig. 4 e Continuous GPS recordings and field labelling of first set of recordings.
width across), which would hinder how the ORP tool created the headlands. Table 1 shows the distance recorded by the GPS during the field operations and the optimised distance calculated by the ORP tool. The table shows as well the excess distance in percentage between the operator's route and the route proposed by the ORP tool. And the area of each labelled field is also included. The operator travelled between 0.2% and 18.4% more distance in each field than the optimised distance calculated by the ORP tool, with an average increase of 8.92%. In total from the twelve fields studied, the ORP tool saved 9.2 km of driving distance of a total of 123 km driven by the operator, i.e. a saving of 7.5%. The route recorded by the machine and the route designed by the ORP tool on field 7 are plotted in Fig. 6. Field 7 has been chosen as an example field, because it is one of the largest fields, and has a total distance difference of 9.3%, which is close to the average distance savings.
Fig. 5 e Continuous GPS recordings and field labelling of second set of recordings.
1 2 3 4 5 6 7 8 9 10 11 12
49,913.42 38,438.32 8838.83 12,230.06 57,057.32 37,247.75 111,498.45 27,956.51 124,360.62 100,486.66 31,674.64 74,576.51
9713.80 8044.20 1850.10 2438.20 11,959.00 8323.40 24,024.00 5202.70 18,259.00 15,811.00 5939.80 11,165.00
9345.52 8005.96 1563.73 2433.87 10,137.22 7773.31 21,984.16 4393.88 17,452.77 14,891.23 5153.83 10,401.60
3.9 0.5 18.3 0.2 18.0 7.1 9.3 18.4 5.6 6.2 15.3 7.3
Histograms showing the number of passes per 5 5 m area in field 7 are depicted in Figs. 7 and 8. The traffic intensity of both the operator's route and the ORP tool during the field operation has been divided into headlands and working area. Table 2 shows data for the trafficking intensity for all twelve fields, distinguishing headlands and working area. The data shows mean and variance values for the number of passes per area in each field. All the fields have a lower mean number of passes per area in both headlands and working area, except for a few were it is equal, with a 95% of confidence interval, for the ORP route compared to the operator's route. The mean values and variances for the whole field were also lower for the optimised route, being up to 0.14 passes per area lower, with the exception of two of the smallest fields where it was equal, with a 95% of confidence interval.
5.
Discussion
The ORP tool reduced the infield travelled distance in all tested fields (see Table 1), with the routes managed by the operator being up to 18.4% longer, saving consequently fuel and hypothetically time during the field operations. Time savings have to be considered hypothetical as the evaluation did not consider speed or acceleration, which might have affected the total time savings of the operation. Including speed in the comparison was not possible, as stated before, because of non-uniform operational speed and breaks taken in the field. Time savings are especially considerable during the connection paths in the headland, as the smoothness of the turn affects directly the speed. Therefore, in order to save time, the optimising algorithm of the ORP tool guides the tractor not to reverse, in order to keep the momentum and avoid extra wheel stress exertion on the soil. However, some studies analysing smooth turning paths in headlands, where speed and acceleration were taken into consideration, showed that reversing was faster than forward turning for distances below 10 m in the x direction (Backman, Piirainen, & Oksanen, 2015; € ben, Meyer zu Helligen, & Schulze Lammers, Sabelhaus, Ro 2013). Even though this size of turning distance, i.e. <10 m, is sometimes modelled by the ORP tool in the optimised routes, the time saved is just of a few seconds per turn, which
154
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Fig. 6 e Plotted operating routes on Field 7 by GPS recordings (left) and by the ORP tool (right).
does not indicate considerable overall time savings. Moreover, not reversing makes the driving easier for the operator. It can therefore be assumed that the ORP tool reduces not only the travelled distance, but also the overall operational time. The operational savings a farmer or agricultural contractor could achieve are large by implementing optimised routes, when it is considered that in a silage growing season a given field may be mowed up to 5 times. Moreover, other operations such as raking and baling could also take advantage of the ORP to find optimised routes. It should also be noted that the comparison is made between the ORP tool and a highly skilled driver, suggesting that if a less skilled driver had been used, the savings gained by using the ORP tool would have been greater. This could create an opportunity for the ORP tool to be used as a training tool for inexperienced drivers, or for a farmer or agricultural contractor to employ a cheaper work force.
The business model employed by agricultural contractors could offer them a unique possibility to gain an advantage using the ORP tool. Currently, farmers pay agricultural contractor to carry out operations on a per area basis, while an agricultural contractor will pay their workers on an hourly basis (Edwards, 2015). If the ORP can be used to reduce the time needed to process an area then the agricultural contractors costs will decrease while their income will remain the same, leading to increased profits. Traffic intensity was also evaluated looking separately into headlands and working areas with the intention of observing how the traffic intensity is distributed in the field (see Table 2). In most fields, the routes proposed by the ORP tool significantly reduced the traffic intensity in the working area, except for three fields, where the number of passes per area was equal, with a 95% of confidence interval. These three fields where no reduction of traffic intensity was made, are some of
Fig. 7 e Histogram showing the number of passes per area in the working area of Field 7.
Fig. 8 e Histogram showing the number of passes per area in the headlands of Field 7.
155
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Table 2 e Recorded and optimised traffic intensity comparison. Working area
Field Field Field Field Field Field Field Field Field Field Field Field
1 2 3 4 5 6 7 8 9 10 11 12
Headlands
GPS mean
GPS var
ORP mean
ORP var
GPS mean
GPS var
ORP mean
ORP var
0.8422 0.8194 0.6484 0.5029 0.8858 0.7912 0.8685 0.4925 0.5184 0.5118 0.4822 0.5646
0.3394 0.3964 0.4502 0.2514 0.3874 0.2973 0.3510 0.2542 0.2699 0.2739 0.3095 0.3322
0.6174 0.7320 0.5641 0.5205 0.8120 0.7574 0.8115 0.5171 0.4936 0.4621 0.5015 0.5256
0.2438 0.2431 0.2575 0.2746 0.1797 0.2060 0.1700 0.2574 0.2500 0.2494 0.2504 0.2495
0.7553 0.7463 0.8455 0.9204 0.8878 1.0606 1.0815 0.9586 0.7206 0.8131 0.9660 0.7783
0.6036 0.8282 1.0672 1.0386 0.9216 1.3674 1.5007 1.1356 0.6025 0.7677 0.9424 0.6135
0.7119 0.8020 0.8351 0.8945 0.8691 1.0927 1.1221 0.8228 0.7281 0.8114 0.9408 0.7633
0.5086 0.6941 0.8417 0.8176 0.8309 1.1618 1.3984 0.8377 0.7266 0.9735 1.1900 0.7264
the smallest fields recorded, which may suggest that the operator and the ORP tool performed equally well, regarding traffic intensity in the working area. In addition, even if the ORP tool confines all the non working distance to the headlands, most of the fields still showed a reduction of mean traffic intensity in the headlands. This displacement of traffic from working area to headlands can be observed for field 7 in the histograms shown in Figs. 7 and 8. It can be seen that the distribution of the number of passes per area is relocated to lower values in the working area, as it assigns all the nonworking distance to the headlands. Consequently the distribution of the number of passes is slightly relocated to higher values in the headlands. In general terms, the routes proposed by the ORP tool for the tested fields had lower means and variances of number of passes per area, meaning lower traffic intensity. An aspect that the traffic intensity evaluation did not consider is the actual wheel traffic. Traffic intensity calculations were made using the GPS position during the operation, so the number of passes per area reflects the area covered by the machine, rather than the compacted area by wheel pressure. It can therefore be observed in the histograms in Figs. 7 and 8, that there are a considerable number of areas with no GPS readings indicating no passes. In general terms, the routes proposed by the ORP tool had lower means and variances in the number of passes per area, meaning lower traffic intensity. This is especially true for the working area, where considerable traffic intensity is reduced, as the ORP tool allocates all non-working distance to the headlands (see Fig. 6). The ORP tool prototype calculates the optimised route considering the headlands along the whole field's borders and not only the top and bottom headlands. This is an interesting feature that many route planning optimisations do not consider (Bakhtiari, Navid, Mehri, & Bochtis, 2011; Bochtis & ~ oz, Vougioukas, 2008; Bochtis et al., 2010; Conesa-Mun Pajares, & Ribeiro, 2016; Jensen et al., 2012). This feature enhances the applicability of optimisation algorithms from theory to practice. The selection of driving direction and division of working area into rows can be a difficult task for operators, especially in large fields. The optimisation algorithm of the ORP tool provides enhanced solutions for the driving direction and row
division, as it is clearly depicted in Fig. 6, where the number of turns inside the working area was visibly reduced. There are however some limitations to the ORP tool. The precision of the tool is dependent on the precision and accuracy of the satellite navigation system. For example, the precision can vary from a few metres, in a cheap GPS receiver, to 1 cm in Real Time Kinematic (RTK) navigation. While this accuracy is independent of the actual ORP tool, it may affect the optimal route calculation. Also, the ORP tool prototype has some existing limitations that need development. Although the tool is designed for NMF operations, the current tool would not be able to optimise ploughing operations, which require a special route pattern in order to fill the empty furrow to the left, by a new furrow during the next pass. Ploughing is the operation that requires highest energy inputs and consequently emits more greenhouse gasses (Dyer & , Petrovic , & Ðevic , 2010), and the Desjardins, 2003; Mileusnic process that causes important subsoil compaction problems (Alakukku et al., 2003). Therefore, this type of operation should be included in the design to fully exploit the possibilities of the tool, and increase sustainability and profitability for the farmer. Fields with narrow areas can present complications for finding an optimised route using the ORP tool. This limitation did not allow finding an optimised route for two of the fields tested which had areas narrower than twice the working width of the mower. Finally, fields with obstacles or considerable slopes will present substantial challenges for the driver to follow the optimised route, which currently limits the applicability of the tool to relatively flat fields with absence or small sized obstacles. Since the ORP tool helps reducing the infield travelled distance and trafficking intensity, it can also be used as a tool for reducing soil compaction. Nevertheless, soil compaction has to be prevented by a combination of different approaches, e.g. increasing soil organic matter (De Neve & Hofman, 2000), adequate crop rotation (Hamza & Anderson, 2005), reducing pressure on soil by low tyre inflation (Alakukku et al., 2003), or working in the field under soil ready conditions (Edwards, , 2016), which can White, Munkholm, Sørensen, & Lamande be combined with the ORP tool. Additional information, such as soil types/moisture, weather, vehicle axle loading, etc., would also need to be made available to the tool so that it could include the consequences of trafficking different areas
156
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
of the field at different times within its optimisation algorithms. The feature of the tool to confine all non-working activity to the headlands makes the prototype a potential tool for controlled traffic farming, as it reduces the traffic in the working area of a field to a minimum. Furthermore, optimised route planners such as the ORP tool, will become indispensable for precision farming in the near future, since they can reduce production costs and operational time, making farming operations more efficient. In general terms, the tool can benefit multiple types of farmers in making their NMF operations more effective, no matter if they practise conventional, organic or precision farming, by reducing the travelled distance in the field by up to 18%. A final advantage of the ORP tool is its development as an application for mobile operating systems, which makes it easy for many farmers to acquire and use it.
6.
Conclusion
The evaluation of the ORP tool prototype was performed comparing travelled distances and traffic intensity by a professional operator during mowing operations and by the optimised route proposed by the tool. In all twelve fields compared, the operator travelled longer distances than the ORP tool's proposed distance, being up to 18.4% longer, with a reduction of 9.2 km (7.5%) in the total infield travelled distance. The tool also generally reduced the traffic intensity, especially in the working area, as it confines all non-working distance to the headlands. Further work is still needed for the ORP tool to be able to fully optimise additional operations, such as ploughing which is the operation with the highest energy requirements and one that significantly affects subsoil compaction. The tool does also present complications when optimising routes for fields with very narrow areas.
references
Alakukku, L. (1996). Persistence of soil compaction due to high axle load traffic. II. Long- term effects on the properties of finetextured and organic soils. Soil and Tillage Research, 37(4), 223e238. http://dx.doi.org/10.1016/0167-1987(96)01017-3. Alakukku, L., Weisskopf, P., Chamen, W. C. T., Tijink, F. G. J., Van Der Linden, J. P., Pires, S., et al. (2003). Prevention strategies for field traffic-induced subsoil compaction: A review part 1. Machine/soil interactions. Soil and Tillage Research, 73(1e2), 145e160. http://dx.doi.org/10.1016/S0167-1987(03)00107-7. Backman, J., Piirainen, P., & Oksanen, T. (2015). Smooth turning path generation for agricultural vehicles in headlands. Biosystems Engineering, 139, 76e86. http://dx.doi.org/10.1016/ j.biosystemseng.2015.08.005. Bakhtiari, A., Navid, H., Mehri, J., & Bochtis, D. D. (2011). Optimal route planning of agricultural field operations using ant colony optimization. Agricultural Engineering International: CIGR Journal, 13(4), 1e16. Retrieved from http://www.cigrjournal. org/index.php/Ejounral/article/view/1939. Bhandral, R., Saggar, S., Bolan, N. S., & Hedley, M. J. (2007). Transformation of nitrogen and nitrous oxide emission from
grassland soils as affected by compaction. Soil and Tillage Research, 94(2), 482e492. http://dx.doi.org/10.1016/ j.still.2006.10.006. Bochtis, D. D., & Sørensen, C. G. (2009). The vehicle routing problem in field logistics part I. Biosystems Engineering, 104(4), 447e457. http://dx.doi.org/10.1016/ j.biosystemseng.2009.09.003. Bochtis, D. D., Sørensen, C. G., Busato, P., & Berruto, R. (2013). Benefits from optimal route planning based on B-patterns. Biosystems Engineering, 115(4), 389e395. http://dx.doi.org/ 10.1016/j.biosystemseng.2013.04.006. Bochtis, D. D., Sørensen, C. G., & Vougioukas, S. G. (2010). Path planning for in-field navigation-aiding of service units. Computers and Electronics in Agriculture, 74(1), 80e90. http:// dx.doi.org/10.1016/j.compag.2010.06.008. Bochtis, D. D., & Vougioukas, S. G. (2008). Minimising the nonworking distance travelled by machines operating in a headland field pattern. Biosystems Engineering, 101(1), 1e12. http://dx.doi.org/10.1016/j.biosystemseng.2008.06.008. ~ oz, J., Pajares, G., & Ribeiro, A. (2016). Mix-opt: A new Conesa-Mun route operator for optimal coverage path planning for a fleet in an agricultural environment. Expert Systems with Applications, 54, 364e378. http://dx.doi.org/10.1016/j.eswa.2015.12.047. De Neve, S., & Hofman, G. (2000). Influence of soil compaction on carbon and nitrogen mineralization of soil organic matter and crop residues. Biology and Fertility of Soils, 30(5e6), 544e549. Dyer, J. A., & Desjardins, R. L. (2003). Simulated farm fieldwork, energy consumption and related greenhouse gas emissions in Canada. Biosystems Engineering, 85(4), 503e513. http:// dx.doi.org/10.1016/S1537-5110(03)00072-2. Edwards, G. T. C. (2015). Field readiness and operation scheduling. Mechanical Engineering. Aarhus University. Edwards, G., White, D. R., Munkholm, L. J., Sørensen, C. G., & , M. (2016). Modelling the readiness of soil for Lamande different methods of tillage. Soil and Tillage Research, 155, 339e350. http://dx.doi.org/10.1016/j.still.2015.08.013. EU. (2016). Short-term outlook for EU arable crops, dairy and meat markets in 2016 and 2017. Retrieved from http://ec.europa.eu/ agriculture/markets-and-prices/short-term-outlook/pdf/20163_en.pdf. FAO. (2015). The state of agricultural commodity markets 2015e16. Trade and food security: Achieving a better balance between national priorities and the collective good (3rd ed.). Rome: Food and Agriculture Organization of the United Nations. Retrieved from http://www.fao.org/3/a-i5090e.pdf. Gła˛b, T. (2014). Effect of soil compaction and N fertilization on soil pore characteristics and physical quality of sandy loam soil under red clover/grass sward. Soil and Tillage Research, 144, 8e19. Hameed, I. A., Bocthis, D. D., & Sorensen, C. G. (2011). Driving angle and track sequence optimization for operational path planning using genetic algorithms. Applied Engineering in Agriculture, 26(6), 1077e1086. http://dx.doi.org/10.13031/ 2013.40615. Hamza, M. A., & Anderson, W. K. (2005). Soil compaction in cropping systems: A review of the nature, causes and possible solutions. Soil and Tillage Research, 82(2), 121e145. http:// dx.doi.org/10.1016/j.still.2004.08.009. ska-Jurkiewicz, A., & van _ Horn, R., Domz_ zał, H., Słowin Ouwerkerk, C. (1995). Soil compaction processes and their effects on the structure of arable soils and the environment. Soil and Tillage Research, 35(1e2), 23e36. http://dx.doi.org/ 10.1016/0167-1987(95)00479-C. Jensen, M. A. F., Bochtis, D., Sorensen, C. G., Blas, M. R., & Lykkegaard, K. L. (2012). In-field and inter-field path planning for agricultural transport units. Computers and Industrial Engineering, 63(4), 1054e1061. http://dx.doi.org/10.1016/ j.cie.2012.07.004.
b i o s y s t e m s e n g i n e e r i n g 1 5 3 ( 2 0 1 7 ) 1 4 9 e1 5 7
Kaivosoja, J., Jackenkroll, M., Linkolehto, R., Weis, M., & Gerhards, R. (2014). Automatic control of farming operations based on spatial web services. Computers and Electronics in Agriculture, 100, 110e115. http://dx.doi.org/10.1016/ j.compag.2013.11.003. Keller, T., Arvidsson, J., Dawidowski, J. B., & Koolen, A. J. (2004). Soil precompression stress: II. A comparison of different compaction tests and stress-displacement behaviour of the soil during wheeling. Soil and Tillage Research, 77(1), 97e108. http://dx.doi.org/10.1016/j.still.2003.11.003. la, F., H Kroulı´k, M., Kumha ula, J., & Honzı´k, I. (2009). The evaluation of agricultural machines field trafficking intensity for different soil tillage technologies. Soil and Tillage Research, 105(1), 171e175. http://dx.doi.org/10.1016/j.still.2009.07.004. Lewis, T. (1998). Evolution of farm management information systems. Computers and Electronics in Agriculture, 19(3), 233e248. http://dx.doi.org/10.1016/S0168-1699(97)00040-9. McHugh, A. D., Tullberg, J. N., & Freebairn, D. M. (2009). Controlled traffic farming restores soil structure. Soil and Tillage Research, 104(1), 164e172. http://dx.doi.org/10.1016/j.still.2008.10.010.
157
, Z. I., Petrovic , D. V., & Ðevic , M. S. (2010). Comparison Mileusnic of tillage systems according to fuel consumption. Energy, 35(1), 221e228. http://dx.doi.org/10.1016/j.energy.2009.09.012. € ben, F., Meyer zu Helligen, L. P., & Schulze Sabelhaus, D., Ro Lammers, P. (2013). Using continuous-curvature paths to generate feasible headland turn manoeuvres. Biosystems Engineering, 116(4), 399e409. http://dx.doi.org/10.1016/ j.biosystemseng.2013.08.012. Sørensen, C. G., & Bochtis, D. D. (2010). Conceptual model of fleet management in agriculture. Biosystems Engineering, 105(1), 41e50. http://dx.doi.org/10.1016/j.biosystemseng.2009.09.009. Tilman, D., Cassman, K. G., Matson, P. A., Naylor, R., & Polasky, S. (2002). Agricultural sustainability and intensive production practices. Nature, 418(6898), 671e677. http://dx.doi.org/ 10.1038/nature01014. Uchida, Y., Clough, T. J., Kelliher, F. M., & Sherlock, R. R. (2008). Effects of aggregate size, soil compaction, and bovine urine on N2O emissions from a pasture soil. Soil Biology and Biochemistry, 40(4), 924e931. http://dx.doi.org/10.1016/ j.soilbio.2007.11.007.