- Email: [email protected]
Effect of control order on steering a simulated underground coal shuttle car
Applied Ergonomics 44 (2013) 225e229
Contents lists available at SciVerse ScienceDirect
Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo
Effect of control order on steering a simulated underground coal shuttle car Robin Burgess-Limerick a, *, Christine Zupanc b, Guy Wallis b a b
Minerals Industry Safety and Health Centre, The University of Queensland, 4072, Australia School of Human Movement Studies, The University of Queensland, 4072, Australia
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 November 2011 Accepted 10 July 2012
Most terrestrial vehicles are steered via a first-order control for vehicle heading, such as a conventional steering wheel. A joystick which provides second-order control of vehicle heading is used to steer some underground coal shuttle cars. A desktop virtual simulation of the situation was employed to compare the steering accuracy of 24 novice participants randomly assigned to either first-order or second-order joystick steering conditions. The average steering accuracy of participants assigned to the first-order joystick condition was superior, however there was considerable individual variability and some participants assigned to the second-order steering condition were able to perform the task equally and successfully. Desktop virtual simulation may be a useful component of training and competency assessment for operators of these vehicles. Ó 2012 Elsevier Ltd and The Ergonomics Society. All rights reserved.
Keywords: Control order Joystick Steering
1. Introduction Shuttle cars are used in underground coal mines to haul coal from a continuous mining machine at the active face “out-bye” to a conveyor system. The cars are of the order of 9 m long and 3.5 m wide, operate in relatively narrow underground roadways (4.5e6.2 m) and are powered via trailing electrical cables. The cars have four wheel steering to provide the manoeuvrability required to turn ninety degree corners. As the name suggests, the cars “shuttle” back and forth between continuous mining machine and conveyor boot end without turning around. Shuttle cars travel at relatively low speeds (up to 10 km/h), carrying heavy loads of coal (of the order of 8 t) over roads which can be very rough. The vehicles are typically operated from a cab located on one out-bye corner of the car. The operator either changes seat with each change of direction to remain facing the direction of travel, or is seated perpendicular to the direction of travel with some degree of seat rotation provided. Shuttle cars are operated in close proximity to pedestrians and have been involved in fatalities and serious injuries (e.g., Burgess-Limerick, 2011a, 2011b; Burgess-Limerick and Steiner, 2006; Hennessy, 2009). The design of most shuttle cars incorporates a directional control-response incompatibility (Burgess-Limerick et al., in press; Zupanc et al., 2007, 2011) which contributes to these risks.
* Corresponding author. Tel.: þ61 7 3346 4084; fax: þ61 7 3346 4067. E-mail address: [email protected] (R. Burgess-Limerick).
Some shuttle cars utilise a steering lever (one-dimensional joystick) to control the direction of travel. Providing a joystick for vehicle steering is potentially problematic. A conventional steering wheel may rotate through multiple revolutions to achieve the maximum steering deviation required and hence the steering sensitivity may be varied through an infinite range. However, the range of motion of a joystick is restricted and the minimum control sensitivity is dictated by the range of steering angle required and the joystick range of motion. A high sensitivity control may be unsuitable, particularly if the vehicle concerned operates on rough roadways because the effects of vehicle motion may feed through the operator’s body to cause unintended forces on the steering control (i.e., “biodynamic feed through” e.g., Gillespie and Sovenyi, 2006; Hill et al., 2008). The joystick provided in some shuttle cars controls the rate of change of steering angle, rather than directly controlling steering angle. Displacing the joystick a constant amount from neutral causes the steering angle to increase at a constant rate. The steering angle continues to change until the steering lever is returned to neutral, and the steering angle then remains constant. Returning the steering to straight ahead requires a second displacement of the steering lever in the opposite direction. In this situation, the maximum steering deviation can be achieved by holding the joystick away from neutral for sufficient time. An advantage of this method of control is that the gain or sensitivity of the control can be adjusted through a large range regardless of the range of throw of the joystick. If the vehicle is moving, holding the joystick in a constant (nonneutral) position causes the rate of change of vehicle heading to
0003-6870/$ e see front matter Ó 2012 Elsevier Ltd and The Ergonomics Society. All rights reserved. http://dx.doi.org/10.1016/j.apergo.2012.07.007
226
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
increase, and this mode of control is consequently termed a second order control, or an acceleration control, for vehicle heading. A joystick which controlled the angle of the steering wheels directly, or a conventional steering wheel, is a first order control of vehicle heading, or velocity control for heading, because a constant displacement of the steering lever causes a constant rate of change of heading. Higher order controls require more steering commands to achieve any given steering manoeuvre (Fig. 1). The study of terrestrial steering control has largely focused on steering wheels and the extent to which operators are able to internalize vehicle dynamics. In particular, research has focused on the distinction between closed-loop vs open-loop control models. Current research has described an unexpected need for appropriately timed visual feedback during basic steering manoeuvres, revealing a surprising misunderstanding of vehicle steering behaviour, even in experienced drivers (Wallis et al., 2002; Cloete and Wallis, 2009). Given the level of naivety of steering dynamics for first order systems and the errors that can occur using this apparently intuitive device, it is important to investigate the possible impact of an inherently more complex second order control system. The effect of control order on error during tracking tasks has been the subject of considerable study, although the paradigms employed have involved abstract laboratory tasks e typically involving controlling the movement of a cursor on a screen. Garvey and Taylor (1959) conducted an experiment in which sixteen participants used a joystick to maintain a cursor centred on a stationary screen position in spite of a continuous series of complex perturbations. Two groups of participants were trained to operate a position control system and an acceleration control system respectively. At the beginning of training, the performance of the position control group was superior, however, this difference was gradually eliminated through training over 23 sessions of ten
Fig. 1. Illustration of the control movements necessary to execute a ninety degree turn with first order and second order controls respectively.
1 min trials. However, when the task was made more difficult by: (i) altering the directional compatibility; (ii) requiring two targets to be tracked with two hands; (iii) tracking the target in two dimensions; or (iv) adding a second visual task; the performance of the acceleration control group deteriorated more than that of the position control group. Similarly, Chernikoff et al. (1960) compared velocity and acceleration control paradigms where six participants controlled the response of a screen cursor via movements of a joystick with the aim of maintaining a stationary cursor position in response to sine wave perturbations of varying amplitude and frequency. The velocity control situation was associated with superior performance. Obermayer et al. (1961) compared the tracking performance of nine participants using a control stick in position, velocity or acceleration control modes. Acceleration control was inferior to both position and rate control in all situations examined. Allen and Jex (1968) examined four pilots performing a tracking tasks with velocity and acceleration control dynamics and reported that acceleration control took longer to learn and was associated with greater variability. Similar conclusions were drawn by Hammerton (1963), Ziegler (1968), Poulton (1969) and Kleinman et al. (1971). Collectively, these results lead Hammerton (1981) to conclude in a review of tracking literature that the superiority of velocity control over higher order systems was clear, and a review by Wickens (1986); see also Fracker and Wickens (1989) observed that error increased markedly between velocity and acceleration control systems, noting “It appears that an increase in RMS tracking error of anywhere from 40 to 100% may be observed with an increase in order from first to second. These differences are quite robust and are maintained across all variations of input, displays, and task loading” (p. 28). More recently, Hancock (1996) observed that tracking error increased with control order for six participants performing a twodimensional task with either trackball or mouse. Similarly, Backs (1997) noted that acceleration control was associated with greater tracking error than velocity control in a task performed by 18 participants. Other related investigations have involved teleoperation control rather than tracking. For example, Kim et al. (1987) examined position and velocity control for 3-axis pick and place tasks using two joysticks. Following extensive training, two subjects completed 80 trials in each condition and it was noted that the position control trials were completed an average of 1.5 times faster. Similarly, Massimino et al. (1989) used a six degree-of-freedom controller and tracking task to compare velocity and acceleration control and concluded that velocity control caused the six participants to make fewer tracking errors that acceleration control. Considering these results, it seems clear that use of a second order joystick to steer the shuttle car will be associated with performance decrements over a first order control system. However, the situations in which the influence of control order on tracking tasks has been examined in the past involved tracking relatively unpredictable perturbations in highly cue reduced laboratory tasks. This situation is not representative of the regular and predictable series of straight sections and ninety degree turns at relatively slow speeds in which joystick steering is utilised when driving an underground shuttle car. Further, while the previous literature suggests that a second-order control will be associated with greater tracking error during the steering task, it is not known whether the magnitudes of the resultant errors are likely to be of practical importance. Consequently, this experiment aimed to determine the effect of first and second-order joystick steering control on a practically relevant measure of steering performance for two groups of novice participants provided with visual cues analogous to the real
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
situation while steering a simulated underground shuttle car with relatively realistic vehicle dynamics travelling through a series of straight roadways and 90 turns. 2. Method The participants’ task involved steering a virtual shuttle car (7.6 m long 3.3 m wide) while it travelled at a constant velocity (6 km/h) through a virtual underground roadway (6.2 m wide). Participants were familiarised with the task by using a steering wheel to navigate the virtual shuttle car for 5 min while observing a plan view of the situation (Fig. 2A). All participants then completed a single 14 min trial using a conventional steering wheel to steer the simulated vehicle through the simulated roadway while provided with a viewpoint analogous to that which would be presented while driving the shuttle car out-bye from active face to conveyor (Fig. 2B). During each trial the virtual shuttle car was driven through a series of 45 m straight sections, interspersed with ten 90 left turns and ten 90 right turns, presented in random order. The participants were instructed to attempt to steer the shuttle car through the roadway without contacting the tunnel walls. Each trial took 14 min to complete. The simulation was created in custom software (based on Cþþ, OpenGL, and the OpenSceneGraph graphics development libraries) and presented on a 22 inch monitor (1280 by 1024 resolution) located 800 mm from the participant. The same apparatus and simulation has also been employed with a separate group of participants to assess the directional control-response compatibility issues associated with this shuttle car design (Burgess-Limerick et al., in press). Repeated measures designs have been criticised as inappropriate for examining the consequences of control order for tracking performance because the transfer between conditions may not be symmetrical (Poulton, 1969, 1974). Consequently, twenty-four
227
novice participants aged 17e26 (18 male) were randomly assigned to either first-order or second-order joystick steering conditions. The threat to this design, given the relatively small number of participants and relatively large individual differences in aptitude for the task, is that random assignment may not result in equivalent groups. Consequently, a trial in which all participants used a first-order steering wheel to complete the task was undertaken to confirm that the randomly assigned groups were equally able to interpret the virtual simulation and perform the task with a conventional steering wheel. Following familiarisation, and completion of the steering wheel trial, all participants completed three trials using a joystick (Saiteck Cyborg Evo) with 20 deg range of motion. The simulator was configured with a dead zone of 1. The vehicle employs a symmetric all-wheel steering mechanism which results in a straightforward relationship between the heading and the angle of the wheels
dH 2vsinq ¼ dt L where H is vehicle heading, v is vehicle velocity (6 km/h), q is average wheel angle (relative to vehicle axis), and L is the vehicle wheelbase (2.7 m). For the 12 participants assigned to the second-order condition, the trials were completed using a joystick configured such that lateral displacement of the joystick beyond the dead zone was proportional to the rate of change of steering angle, up to a maximum rate of change of steering angle of 12 /s. The steering dynamics of the virtual shuttle car approximated those of a physical shuttle car. In this condition, for ‘dead zone’ angle a ¼ 1, qmax ¼ 21.5 , joystick angle f(fmax ¼ 10 ), f > a and q < qmax:
dq ðf aÞ ¼ 4þ8 dt ðfmax aÞ For movements of the joystick in the opposite direction (f < 0) symmetric equations apply with a dead zone extending to a . For the 12 participants assigned to the first-order condition, the joystick was configured such that the steering displacement of the vehicle was proportional to the lateral displacement of the joystick beyond the dead zone. That is, in the first order condition, for f > a:
q ¼ qmax
ðf aÞ ðfmax aÞ
under the constraint that if q is changing, it changes at a fixed rate of 12 /s. This rate of change of steering displacement created a slight lag in response to large joystick movements which corresponds to the lag experienced in the hydraulics of a shuttle car. The time during which any part of the simulated vehicle was in contact with the wall of the roadway was calculated as a dependent measure which provides a meaningful measure of the consequences of steering error. A mixed two-way (joystick condition trial) ANOVA with trial treated as a repeated measure was employed to analyse these data.
3. Results & discussion
Fig. 2. The top-down view presented of the simulated task (A) provided for the familiarisation and driver’s view, (B) presented for subsequent trials.
Fig. 3 presents the mean (95% confidence intervals) duration of contact between simulated shuttle car and roadway wall. Random allocation of participants was successful in achieving two groups of participants which did not differ in their ability to perform the task during trial 0 in which the task was performed with a conventional steering wheel. Two-way mixed ANOVA conducted on the data
228
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
Fig. 3. Mean (95% confidence interval) time for which the virtual vehicle was in contact with the virtual roadway roadway walls during each 14 min trial.
for trials 1e3 revealed significant main effects of joystick condition (F[1,22] ¼ 12.37, p ¼ 0.0019) and trial (F[2,22] ¼ 19.58, p < 0.0001) and a significant interaction between trial and joystick condition (F[2,22] ¼ 7.313, p ¼ 0.0018), suggesting that steering performance as measured by the time the shuttle car was in contact with the tunnel walls was superior using a first order joystick, but that the difference in performance was reduced after three 14 min trials. The results are consistent with those from previous cursor tracking tasks (e.g., Allen and Jex, 1968; Backs, 1997; Chernikoff et al., 1960; Garvey and Taylor, 1959; Hancock, 1996; Obermayer et al., 1961) in suggesting that higher order control systems are associated with less accurate steering performance; and that such performance differences reduce with practice. The duration of practice required to eliminate the performance differences is unknown. A potential confound in the experimental design exists in that performing a trial with the first order steering wheel may have improved the subsequent performance of participants assigned to the first order joystick more than the performance of participants assigned to the second order joystick group. However, the magnitude of any additional benefit for the first-order joystick group is likely small. All participants had experience driving a motor vehicle, and thus all participants had considerable experience with
first-order control of a vehicle via a steering wheel. Further, Poulton (1969) highlighted data which demonstrate that prior practice with first order control also has a positive benefit on performance with a second order control. Fig. 4 presents individual data points for collision time for each trial. These data highlight individual differences in participants’ responses to the second order joystick steering condition. While, on average, the performance of participants allocated to the secondorder condition was less accurate than those participants allocated to the first order steering condition, at least 1/3 of the participants in the second-order condition exhibited steering performance which was practically indistinguishable from those participants allocated to the first order condition. An alternate steering mechanism for shuttle cars may be preferable e a first-order joystick may be suitable if biodynamic feed through and related sensitivity issues can be overcome, perhaps by optimising the response delay; or a conventional steering wheel may be preferable. Steering wheels are currently employed on shuttle cars used in Australian coal mines, however most designs also have a known directional incompatibility issue while driving the shuttle car towards the continuous miner (Zupanc et al., 2007, 2011). One shuttle car model is available in Australia which overcomes these issues by providing a steering wheel and seat which rotates to always face the direction of travel and maintain directional compatibility, however this is not in widespread use. Where second-order joystick steering continues to be employed to control shuttle cars, the implications of these data are that sufficient training to achieve competence is required, and that the duration of training required may differ considerably across people. Further, this implies that an objective method of determining competence is required. Desktop virtual reality simulation of the shuttle car may provide a potential component of both training and competence assessment in this situation (Tichon and BurgessLimerick, 2011).
Acknowledgement Funding for this project was received from Sandvik Mining & Construction Tomago Pty Ltd. (SMC). SMC did not have input into study design; the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.
References
Fig. 4. Time for which the virtual vehicle was in contact with the virtual roadway roadway walls during each 14 min trial for each trial completed by each participant.
Allen, R.W., Jex, H.R., 1968. An Experimental Investigation of Compensatory and Pursuit Tracking Displays with Rate and Acceleration Control Dynamics and a Disturbance Input. NASA, Washington D.C (NASA CR-1082). Backs, R.W., 1997. Psychophysiological aspects of selective and divided attention during continuous manual tracking. Acta Psychologica 96, 167e191. Burgess-Limerick, R., 2011a. Avoiding collisions in underground mines. Ergonomics Australia e HFESA 2011 Conference Edition 11, 7. Burgess-Limerick, R., 2011b. Injuries associated with underground coal mining equipment. The Ergonomics Open Journal 4 (Suppl. 2eM1), 62e73. Burgess-Limerick, R., Steiner, L., 2006. Injuries associated with continuous miners, shuttle cars, load- haul-dump, and personnel transport in new South Wales underground coal mines. Mining Technology (TIMM A) 115, 160e168. Burgess-Limerick, R., Zupanc, C., Wallis, G. Directional control-response compatibility of joystick steered shuttle cars. Ergonomics, in press. Chernikoff, R., Duey, J.W., Taylor, F.V., 1960. Two-dimensional tracking with identical and different control dynamics in each coordinate. Journal of Experimental Psychology 60, 318e322. Cloete, S., Wallis, G., 2009. Limitations of feedforward control in multiple-phase steering movements. Experimental Brain Research 195, 481e487. Fracker, M.L., Wickens, C.D., 1989. Resources, confusions, and compatibility in dualaxis tracking: displays, controls, and dynamics. Journal of Experimental Psychology: Human Perception and Performance. 15, 80e96. Garvey, W.D., Taylor, F.V., 1959. Interactions among operator variables, system dynamics, and task-induced stress. Journal of Applied Psychology 43, 79e85.
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229 Gillespie, R.B., Sovenyi, S., 2006. Model-based cancellation of biodynamic feedthrough using a force-reflecting joystick. Journal of Dynamic Systems, Measurement, and Control. 128, 94e103. Hammerton, M., 1963. Retention of learning in a difficult tracking task. Journal of Experimental Psychology 66, 108e110. Hammerton, M.,1981. Tracking. In: Holding, D.H. (Ed.), Human Skills. J. Wiley, New York. Hancock, P.A., 1996. Effects of control order, augmented feedback, input device and practice on tracking performance and perceived workload. Ergonomics 39, 1146e1162. Hennessy, A., 2009. Inquest into the Death of Jason George Elliott Blee. Office of the State Coroner. ROCK-COR-42/2007. http://www.courts.qld.gov.au/BleeJGE20090910.pdf. Hill, S.G., Metcalfe, J.S., McDowell, K., 2008. The Use of a Steering Shaping Function to Improve Human Performance in By-Wire Vehicles. Army Research Laboratory, Aberdeen MD. Report ARL-TR-4387. Kim, W.S., Tendick, F., Ellis, S.P., Stark, L.W., 1987. A comparison of position and rate control for telemanipulations with consideration of manipulator system dynamics. IEEE Journal of Robotics and Automation RA-3, 426e436. Kleinman, D.L., Baron, S., Levison, W.H., 1971. A control theoretic approach to manned-vehicle systems analysis. IEEE Transactions on Automatic Control AC16, 824e832. Massimino, M.J., Sheridan, T.B., Roseborough, J.B., 1989. One hand tracking in six degree of freedom. Proceedings of IEEE International Conference on Systems, Man, and Cybernetics, 498e503.
229
Obermayer, R.W., Swartz, W.F., Muckler, F.A., 1961. The interaction of information displays with control system dynamics in continuous tracking. Journal of Applied Psychology 45, 369e375. Poulton, E.C., 1969. Bias in experimental comparisons between equipments due to the order of testing. Ergonomics 12, 675e687. Poulton, E.C., 1974. Tracking Skill and Manual Control. Academic Press, New York. Tichon, J., Burgess-Limerick, R., 2011. A review of virtual reality as a medium for safety related training in mining. Journal of Health & Safety Research & Practice 3, 33e40. Wallis, G., Chatziastros, A., Bülthoff, H.H., 2002. An unexpected role for visual feedback in vehicle steering control. Current Biology 12, 295e299. Wickens, C.D., 1986. The effects of control dynamics on performance. In: Boff, K.R., Kaufman, L., Thomas, J.P. (Eds.), Handbook of Perception and Human Performance. Cognitive Processes and Performance, vol. II. Wiley, New York, pp. 39-1e39-60. Ziegler, P.M., 1968. Single and dual axis tracking as a function of system dynamics. Human Factors 10, 273e276. Zupanc, C., Burgess-Limerick, R., Wallis, G., 2007. Performance as a consequence of alternating control-response compatibility: evidence from a coal mine shuttle car simulator. Human Factors 49, 628e636. Zupanc, C., Burgess-Limerick, R., Wallis, G., 2011. Effect of age on learning to drive a virtual coal mine shuttle car. The Ergonomics Open Journal 4 (Suppl. 2-M1), 112e124.
Contents lists available at SciVerse ScienceDirect
Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo
Effect of control order on steering a simulated underground coal shuttle car Robin Burgess-Limerick a, *, Christine Zupanc b, Guy Wallis b a b
Minerals Industry Safety and Health Centre, The University of Queensland, 4072, Australia School of Human Movement Studies, The University of Queensland, 4072, Australia
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 November 2011 Accepted 10 July 2012
Most terrestrial vehicles are steered via a first-order control for vehicle heading, such as a conventional steering wheel. A joystick which provides second-order control of vehicle heading is used to steer some underground coal shuttle cars. A desktop virtual simulation of the situation was employed to compare the steering accuracy of 24 novice participants randomly assigned to either first-order or second-order joystick steering conditions. The average steering accuracy of participants assigned to the first-order joystick condition was superior, however there was considerable individual variability and some participants assigned to the second-order steering condition were able to perform the task equally and successfully. Desktop virtual simulation may be a useful component of training and competency assessment for operators of these vehicles. Ó 2012 Elsevier Ltd and The Ergonomics Society. All rights reserved.
Keywords: Control order Joystick Steering
1. Introduction Shuttle cars are used in underground coal mines to haul coal from a continuous mining machine at the active face “out-bye” to a conveyor system. The cars are of the order of 9 m long and 3.5 m wide, operate in relatively narrow underground roadways (4.5e6.2 m) and are powered via trailing electrical cables. The cars have four wheel steering to provide the manoeuvrability required to turn ninety degree corners. As the name suggests, the cars “shuttle” back and forth between continuous mining machine and conveyor boot end without turning around. Shuttle cars travel at relatively low speeds (up to 10 km/h), carrying heavy loads of coal (of the order of 8 t) over roads which can be very rough. The vehicles are typically operated from a cab located on one out-bye corner of the car. The operator either changes seat with each change of direction to remain facing the direction of travel, or is seated perpendicular to the direction of travel with some degree of seat rotation provided. Shuttle cars are operated in close proximity to pedestrians and have been involved in fatalities and serious injuries (e.g., Burgess-Limerick, 2011a, 2011b; Burgess-Limerick and Steiner, 2006; Hennessy, 2009). The design of most shuttle cars incorporates a directional control-response incompatibility (Burgess-Limerick et al., in press; Zupanc et al., 2007, 2011) which contributes to these risks.
* Corresponding author. Tel.: þ61 7 3346 4084; fax: þ61 7 3346 4067. E-mail address: [email protected] (R. Burgess-Limerick).
Some shuttle cars utilise a steering lever (one-dimensional joystick) to control the direction of travel. Providing a joystick for vehicle steering is potentially problematic. A conventional steering wheel may rotate through multiple revolutions to achieve the maximum steering deviation required and hence the steering sensitivity may be varied through an infinite range. However, the range of motion of a joystick is restricted and the minimum control sensitivity is dictated by the range of steering angle required and the joystick range of motion. A high sensitivity control may be unsuitable, particularly if the vehicle concerned operates on rough roadways because the effects of vehicle motion may feed through the operator’s body to cause unintended forces on the steering control (i.e., “biodynamic feed through” e.g., Gillespie and Sovenyi, 2006; Hill et al., 2008). The joystick provided in some shuttle cars controls the rate of change of steering angle, rather than directly controlling steering angle. Displacing the joystick a constant amount from neutral causes the steering angle to increase at a constant rate. The steering angle continues to change until the steering lever is returned to neutral, and the steering angle then remains constant. Returning the steering to straight ahead requires a second displacement of the steering lever in the opposite direction. In this situation, the maximum steering deviation can be achieved by holding the joystick away from neutral for sufficient time. An advantage of this method of control is that the gain or sensitivity of the control can be adjusted through a large range regardless of the range of throw of the joystick. If the vehicle is moving, holding the joystick in a constant (nonneutral) position causes the rate of change of vehicle heading to
0003-6870/$ e see front matter Ó 2012 Elsevier Ltd and The Ergonomics Society. All rights reserved. http://dx.doi.org/10.1016/j.apergo.2012.07.007
226
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
increase, and this mode of control is consequently termed a second order control, or an acceleration control, for vehicle heading. A joystick which controlled the angle of the steering wheels directly, or a conventional steering wheel, is a first order control of vehicle heading, or velocity control for heading, because a constant displacement of the steering lever causes a constant rate of change of heading. Higher order controls require more steering commands to achieve any given steering manoeuvre (Fig. 1). The study of terrestrial steering control has largely focused on steering wheels and the extent to which operators are able to internalize vehicle dynamics. In particular, research has focused on the distinction between closed-loop vs open-loop control models. Current research has described an unexpected need for appropriately timed visual feedback during basic steering manoeuvres, revealing a surprising misunderstanding of vehicle steering behaviour, even in experienced drivers (Wallis et al., 2002; Cloete and Wallis, 2009). Given the level of naivety of steering dynamics for first order systems and the errors that can occur using this apparently intuitive device, it is important to investigate the possible impact of an inherently more complex second order control system. The effect of control order on error during tracking tasks has been the subject of considerable study, although the paradigms employed have involved abstract laboratory tasks e typically involving controlling the movement of a cursor on a screen. Garvey and Taylor (1959) conducted an experiment in which sixteen participants used a joystick to maintain a cursor centred on a stationary screen position in spite of a continuous series of complex perturbations. Two groups of participants were trained to operate a position control system and an acceleration control system respectively. At the beginning of training, the performance of the position control group was superior, however, this difference was gradually eliminated through training over 23 sessions of ten
Fig. 1. Illustration of the control movements necessary to execute a ninety degree turn with first order and second order controls respectively.
1 min trials. However, when the task was made more difficult by: (i) altering the directional compatibility; (ii) requiring two targets to be tracked with two hands; (iii) tracking the target in two dimensions; or (iv) adding a second visual task; the performance of the acceleration control group deteriorated more than that of the position control group. Similarly, Chernikoff et al. (1960) compared velocity and acceleration control paradigms where six participants controlled the response of a screen cursor via movements of a joystick with the aim of maintaining a stationary cursor position in response to sine wave perturbations of varying amplitude and frequency. The velocity control situation was associated with superior performance. Obermayer et al. (1961) compared the tracking performance of nine participants using a control stick in position, velocity or acceleration control modes. Acceleration control was inferior to both position and rate control in all situations examined. Allen and Jex (1968) examined four pilots performing a tracking tasks with velocity and acceleration control dynamics and reported that acceleration control took longer to learn and was associated with greater variability. Similar conclusions were drawn by Hammerton (1963), Ziegler (1968), Poulton (1969) and Kleinman et al. (1971). Collectively, these results lead Hammerton (1981) to conclude in a review of tracking literature that the superiority of velocity control over higher order systems was clear, and a review by Wickens (1986); see also Fracker and Wickens (1989) observed that error increased markedly between velocity and acceleration control systems, noting “It appears that an increase in RMS tracking error of anywhere from 40 to 100% may be observed with an increase in order from first to second. These differences are quite robust and are maintained across all variations of input, displays, and task loading” (p. 28). More recently, Hancock (1996) observed that tracking error increased with control order for six participants performing a twodimensional task with either trackball or mouse. Similarly, Backs (1997) noted that acceleration control was associated with greater tracking error than velocity control in a task performed by 18 participants. Other related investigations have involved teleoperation control rather than tracking. For example, Kim et al. (1987) examined position and velocity control for 3-axis pick and place tasks using two joysticks. Following extensive training, two subjects completed 80 trials in each condition and it was noted that the position control trials were completed an average of 1.5 times faster. Similarly, Massimino et al. (1989) used a six degree-of-freedom controller and tracking task to compare velocity and acceleration control and concluded that velocity control caused the six participants to make fewer tracking errors that acceleration control. Considering these results, it seems clear that use of a second order joystick to steer the shuttle car will be associated with performance decrements over a first order control system. However, the situations in which the influence of control order on tracking tasks has been examined in the past involved tracking relatively unpredictable perturbations in highly cue reduced laboratory tasks. This situation is not representative of the regular and predictable series of straight sections and ninety degree turns at relatively slow speeds in which joystick steering is utilised when driving an underground shuttle car. Further, while the previous literature suggests that a second-order control will be associated with greater tracking error during the steering task, it is not known whether the magnitudes of the resultant errors are likely to be of practical importance. Consequently, this experiment aimed to determine the effect of first and second-order joystick steering control on a practically relevant measure of steering performance for two groups of novice participants provided with visual cues analogous to the real
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
situation while steering a simulated underground shuttle car with relatively realistic vehicle dynamics travelling through a series of straight roadways and 90 turns. 2. Method The participants’ task involved steering a virtual shuttle car (7.6 m long 3.3 m wide) while it travelled at a constant velocity (6 km/h) through a virtual underground roadway (6.2 m wide). Participants were familiarised with the task by using a steering wheel to navigate the virtual shuttle car for 5 min while observing a plan view of the situation (Fig. 2A). All participants then completed a single 14 min trial using a conventional steering wheel to steer the simulated vehicle through the simulated roadway while provided with a viewpoint analogous to that which would be presented while driving the shuttle car out-bye from active face to conveyor (Fig. 2B). During each trial the virtual shuttle car was driven through a series of 45 m straight sections, interspersed with ten 90 left turns and ten 90 right turns, presented in random order. The participants were instructed to attempt to steer the shuttle car through the roadway without contacting the tunnel walls. Each trial took 14 min to complete. The simulation was created in custom software (based on Cþþ, OpenGL, and the OpenSceneGraph graphics development libraries) and presented on a 22 inch monitor (1280 by 1024 resolution) located 800 mm from the participant. The same apparatus and simulation has also been employed with a separate group of participants to assess the directional control-response compatibility issues associated with this shuttle car design (Burgess-Limerick et al., in press). Repeated measures designs have been criticised as inappropriate for examining the consequences of control order for tracking performance because the transfer between conditions may not be symmetrical (Poulton, 1969, 1974). Consequently, twenty-four
227
novice participants aged 17e26 (18 male) were randomly assigned to either first-order or second-order joystick steering conditions. The threat to this design, given the relatively small number of participants and relatively large individual differences in aptitude for the task, is that random assignment may not result in equivalent groups. Consequently, a trial in which all participants used a first-order steering wheel to complete the task was undertaken to confirm that the randomly assigned groups were equally able to interpret the virtual simulation and perform the task with a conventional steering wheel. Following familiarisation, and completion of the steering wheel trial, all participants completed three trials using a joystick (Saiteck Cyborg Evo) with 20 deg range of motion. The simulator was configured with a dead zone of 1. The vehicle employs a symmetric all-wheel steering mechanism which results in a straightforward relationship between the heading and the angle of the wheels
dH 2vsinq ¼ dt L where H is vehicle heading, v is vehicle velocity (6 km/h), q is average wheel angle (relative to vehicle axis), and L is the vehicle wheelbase (2.7 m). For the 12 participants assigned to the second-order condition, the trials were completed using a joystick configured such that lateral displacement of the joystick beyond the dead zone was proportional to the rate of change of steering angle, up to a maximum rate of change of steering angle of 12 /s. The steering dynamics of the virtual shuttle car approximated those of a physical shuttle car. In this condition, for ‘dead zone’ angle a ¼ 1, qmax ¼ 21.5 , joystick angle f(fmax ¼ 10 ), f > a and q < qmax:
dq ðf aÞ ¼ 4þ8 dt ðfmax aÞ For movements of the joystick in the opposite direction (f < 0) symmetric equations apply with a dead zone extending to a . For the 12 participants assigned to the first-order condition, the joystick was configured such that the steering displacement of the vehicle was proportional to the lateral displacement of the joystick beyond the dead zone. That is, in the first order condition, for f > a:
q ¼ qmax
ðf aÞ ðfmax aÞ
under the constraint that if q is changing, it changes at a fixed rate of 12 /s. This rate of change of steering displacement created a slight lag in response to large joystick movements which corresponds to the lag experienced in the hydraulics of a shuttle car. The time during which any part of the simulated vehicle was in contact with the wall of the roadway was calculated as a dependent measure which provides a meaningful measure of the consequences of steering error. A mixed two-way (joystick condition trial) ANOVA with trial treated as a repeated measure was employed to analyse these data.
3. Results & discussion
Fig. 2. The top-down view presented of the simulated task (A) provided for the familiarisation and driver’s view, (B) presented for subsequent trials.
Fig. 3 presents the mean (95% confidence intervals) duration of contact between simulated shuttle car and roadway wall. Random allocation of participants was successful in achieving two groups of participants which did not differ in their ability to perform the task during trial 0 in which the task was performed with a conventional steering wheel. Two-way mixed ANOVA conducted on the data
228
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229
Fig. 3. Mean (95% confidence interval) time for which the virtual vehicle was in contact with the virtual roadway roadway walls during each 14 min trial.
for trials 1e3 revealed significant main effects of joystick condition (F[1,22] ¼ 12.37, p ¼ 0.0019) and trial (F[2,22] ¼ 19.58, p < 0.0001) and a significant interaction between trial and joystick condition (F[2,22] ¼ 7.313, p ¼ 0.0018), suggesting that steering performance as measured by the time the shuttle car was in contact with the tunnel walls was superior using a first order joystick, but that the difference in performance was reduced after three 14 min trials. The results are consistent with those from previous cursor tracking tasks (e.g., Allen and Jex, 1968; Backs, 1997; Chernikoff et al., 1960; Garvey and Taylor, 1959; Hancock, 1996; Obermayer et al., 1961) in suggesting that higher order control systems are associated with less accurate steering performance; and that such performance differences reduce with practice. The duration of practice required to eliminate the performance differences is unknown. A potential confound in the experimental design exists in that performing a trial with the first order steering wheel may have improved the subsequent performance of participants assigned to the first order joystick more than the performance of participants assigned to the second order joystick group. However, the magnitude of any additional benefit for the first-order joystick group is likely small. All participants had experience driving a motor vehicle, and thus all participants had considerable experience with
first-order control of a vehicle via a steering wheel. Further, Poulton (1969) highlighted data which demonstrate that prior practice with first order control also has a positive benefit on performance with a second order control. Fig. 4 presents individual data points for collision time for each trial. These data highlight individual differences in participants’ responses to the second order joystick steering condition. While, on average, the performance of participants allocated to the secondorder condition was less accurate than those participants allocated to the first order steering condition, at least 1/3 of the participants in the second-order condition exhibited steering performance which was practically indistinguishable from those participants allocated to the first order condition. An alternate steering mechanism for shuttle cars may be preferable e a first-order joystick may be suitable if biodynamic feed through and related sensitivity issues can be overcome, perhaps by optimising the response delay; or a conventional steering wheel may be preferable. Steering wheels are currently employed on shuttle cars used in Australian coal mines, however most designs also have a known directional incompatibility issue while driving the shuttle car towards the continuous miner (Zupanc et al., 2007, 2011). One shuttle car model is available in Australia which overcomes these issues by providing a steering wheel and seat which rotates to always face the direction of travel and maintain directional compatibility, however this is not in widespread use. Where second-order joystick steering continues to be employed to control shuttle cars, the implications of these data are that sufficient training to achieve competence is required, and that the duration of training required may differ considerably across people. Further, this implies that an objective method of determining competence is required. Desktop virtual reality simulation of the shuttle car may provide a potential component of both training and competence assessment in this situation (Tichon and BurgessLimerick, 2011).
Acknowledgement Funding for this project was received from Sandvik Mining & Construction Tomago Pty Ltd. (SMC). SMC did not have input into study design; the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.
References
Fig. 4. Time for which the virtual vehicle was in contact with the virtual roadway roadway walls during each 14 min trial for each trial completed by each participant.
Allen, R.W., Jex, H.R., 1968. An Experimental Investigation of Compensatory and Pursuit Tracking Displays with Rate and Acceleration Control Dynamics and a Disturbance Input. NASA, Washington D.C (NASA CR-1082). Backs, R.W., 1997. Psychophysiological aspects of selective and divided attention during continuous manual tracking. Acta Psychologica 96, 167e191. Burgess-Limerick, R., 2011a. Avoiding collisions in underground mines. Ergonomics Australia e HFESA 2011 Conference Edition 11, 7. Burgess-Limerick, R., 2011b. Injuries associated with underground coal mining equipment. The Ergonomics Open Journal 4 (Suppl. 2eM1), 62e73. Burgess-Limerick, R., Steiner, L., 2006. Injuries associated with continuous miners, shuttle cars, load- haul-dump, and personnel transport in new South Wales underground coal mines. Mining Technology (TIMM A) 115, 160e168. Burgess-Limerick, R., Zupanc, C., Wallis, G. Directional control-response compatibility of joystick steered shuttle cars. Ergonomics, in press. Chernikoff, R., Duey, J.W., Taylor, F.V., 1960. Two-dimensional tracking with identical and different control dynamics in each coordinate. Journal of Experimental Psychology 60, 318e322. Cloete, S., Wallis, G., 2009. Limitations of feedforward control in multiple-phase steering movements. Experimental Brain Research 195, 481e487. Fracker, M.L., Wickens, C.D., 1989. Resources, confusions, and compatibility in dualaxis tracking: displays, controls, and dynamics. Journal of Experimental Psychology: Human Perception and Performance. 15, 80e96. Garvey, W.D., Taylor, F.V., 1959. Interactions among operator variables, system dynamics, and task-induced stress. Journal of Applied Psychology 43, 79e85.
R. Burgess-Limerick et al. / Applied Ergonomics 44 (2013) 225e229 Gillespie, R.B., Sovenyi, S., 2006. Model-based cancellation of biodynamic feedthrough using a force-reflecting joystick. Journal of Dynamic Systems, Measurement, and Control. 128, 94e103. Hammerton, M., 1963. Retention of learning in a difficult tracking task. Journal of Experimental Psychology 66, 108e110. Hammerton, M.,1981. Tracking. In: Holding, D.H. (Ed.), Human Skills. J. Wiley, New York. Hancock, P.A., 1996. Effects of control order, augmented feedback, input device and practice on tracking performance and perceived workload. Ergonomics 39, 1146e1162. Hennessy, A., 2009. Inquest into the Death of Jason George Elliott Blee. Office of the State Coroner. ROCK-COR-42/2007. http://www.courts.qld.gov.au/BleeJGE20090910.pdf. Hill, S.G., Metcalfe, J.S., McDowell, K., 2008. The Use of a Steering Shaping Function to Improve Human Performance in By-Wire Vehicles. Army Research Laboratory, Aberdeen MD. Report ARL-TR-4387. Kim, W.S., Tendick, F., Ellis, S.P., Stark, L.W., 1987. A comparison of position and rate control for telemanipulations with consideration of manipulator system dynamics. IEEE Journal of Robotics and Automation RA-3, 426e436. Kleinman, D.L., Baron, S., Levison, W.H., 1971. A control theoretic approach to manned-vehicle systems analysis. IEEE Transactions on Automatic Control AC16, 824e832. Massimino, M.J., Sheridan, T.B., Roseborough, J.B., 1989. One hand tracking in six degree of freedom. Proceedings of IEEE International Conference on Systems, Man, and Cybernetics, 498e503.
229
Obermayer, R.W., Swartz, W.F., Muckler, F.A., 1961. The interaction of information displays with control system dynamics in continuous tracking. Journal of Applied Psychology 45, 369e375. Poulton, E.C., 1969. Bias in experimental comparisons between equipments due to the order of testing. Ergonomics 12, 675e687. Poulton, E.C., 1974. Tracking Skill and Manual Control. Academic Press, New York. Tichon, J., Burgess-Limerick, R., 2011. A review of virtual reality as a medium for safety related training in mining. Journal of Health & Safety Research & Practice 3, 33e40. Wallis, G., Chatziastros, A., Bülthoff, H.H., 2002. An unexpected role for visual feedback in vehicle steering control. Current Biology 12, 295e299. Wickens, C.D., 1986. The effects of control dynamics on performance. In: Boff, K.R., Kaufman, L., Thomas, J.P. (Eds.), Handbook of Perception and Human Performance. Cognitive Processes and Performance, vol. II. Wiley, New York, pp. 39-1e39-60. Ziegler, P.M., 1968. Single and dual axis tracking as a function of system dynamics. Human Factors 10, 273e276. Zupanc, C., Burgess-Limerick, R., Wallis, G., 2007. Performance as a consequence of alternating control-response compatibility: evidence from a coal mine shuttle car simulator. Human Factors 49, 628e636. Zupanc, C., Burgess-Limerick, R., Wallis, G., 2011. Effect of age on learning to drive a virtual coal mine shuttle car. The Ergonomics Open Journal 4 (Suppl. 2-M1), 112e124.