CHAPTER EIGHT
Final Remarks Nowadays, bio-inspired algorithms have become a very important tool to solve engineering problems. Given that using classic optimization algorithms often implies computing and iteratively evaluating Hessians and gradients, the computational cost of this kind of algorithms—in some cases—may be excessively high. On the other hand, calculations of bioinspired algorithms involve simple operations, most of them being linear equations. Furthermore, bio-inspired algorithms are easy to parallelize using a “divide and conquer” strategy, in which each individual in the population of the algorithm can be seen as a subprocess of the whole bio-inspired system. That means, each individual can be treated as a thread to solve a part of the problem because each individual of the population can be evaluated independently. If required, a phase of communication between individuals could be performed. So, using bio-inspired algorithms the computation complexity and execution time to solve engineering problems are often reduced, making them an efficient alternative to solve these kinds of problems. It is important to mention that bio-inspired algorithms have a very explicit and easy-to-understand operation because they follow a simplified model of a biological entity. This makes their programming produce explicit codes and versions; handling change is also made simple. The purpose of this chapter is to present a book summary with final remarks and general conclusions, and to offer a forecast regarding future work. Several state-of-the-art applications to solve engineering problems, ranging from intelligent pattern recognition, object reconstruction, robot control and vision, intelligent identification, and control of nonlinear systems have been presented in this book. All of the applications presented are characterized by explicit, efficient, and robust systems, which involve using bio-inspired algorithms to function. As a result, the authors believe that using and designing bio-inspired algorithms is, and will, continue to be a strong and growing research area in engineering, and they hope that this book provides young researchers in this field inspiration to continue producing engineering solutions. Chapter 1 offered a brief review of bio-inspiring algorithms and their importance to solving complex optimization problems in engineering. Bio-inspired Algorithms for Engineering https://doi.org/10.1016/B978-0-12-813788-8.00008-1
© 2018 Elsevier Inc.
All rights reserved.
129
130
Bio-inspired Algorithms for Engineering
In addition, Chapter 1 briefly introduced bio-inspired algorithms used throughout the rest of the book. These included PSO, ABC, μABC, DE, and BFO. Chapter 2 presented an approach to perform large-scale data classification using SVMs trained using KA combined with ABC, μABC, DE, and PSO. The combination of KA and bio-inspired algorithms allowed us to obtain a parallelized system of classification that was effectively applied to solve pattern recognition with data of large dimension, keeping its computational complexity lower than previous large-scale classifiers that use SVM. Our proposed SVM-bio-inspired algorithm can be used to solve problems as classification of chromosomes, spam filtering, information security, and other problems where the dimension and/or amount of data is very large and where the generalization of knowledge is highly desired to obtain a low training error. Chapter 3 showed the design and implementation of a method to reconstruct 3D surfaces from point-clouds using RBF neural networks, which have been adjusted using PSO. Meshing functions in order to interpolate point-clouds to obtain compact surface representations is very important for CAD, robot mapping, and object description and recognition among other important applications. The results obtained using our algorithm show that although the obtained surfaces were not continuous, they can be used as compact descriptors for pattern recognition process and environmental mapping. This is because our proposal is fast enough to be implemented in real time, and the reduction of the number of parameters used to describe a shape with 3D point clouds is significant. Chapter 4 presented an application of soft computing algorithms, such as PSO to solve important computer vision problems like image tracking and plane detection. These problem-solving capacities are very important since they can be used to perform more complex tasks, such as grasping and robot navigation for detection of obstacles and mapping. The presented image tracking algorithm is capable of working with large images. This is as a result of our having used an image pyramid search and PSO for looking for the best template match that improves the speed of the template search. The plane detection algorithm also used PSO with data obtained from an RGBD sensor and used it to conduct a search in the image plane. However, complementing the data on the plane with depth information to construct planes allowed us to implement a robust algorithm for noise and outliers. Chapter 5 dealt with a soft computing approach that is able to avoid obstacles and to move a robot to reach a goal. The approach is based on
Final Remarks
131
the PSO algorithm, where each particle represents a potential solution of a new position for the robot. Once the best particle of the actual iteration is selected, the robot is moved to the position that the best particle represents. This algorithm was tested with nonholonomic and holonomic robots, and it proved that bio-inspired algorithms are able to solve local navigation problems in real-time. Chapter 6 presented the application of bio-inspired algorithms to improve neural identifiers for discrete-time unknown nonlinear systems. PSO was used to improve two kinds of neural identifiers. First, PSO was used to identify conditions conducive for an EKF learning algorithm to train a RHONN. The purpose of this undertaking was to identify a dynamic mathematical model to serve as a linear induction motor benchmark; second, the same enhanced PSO-EKF was used to train a recurrent multilayer perceptron in order to obtain an accurate neural model for forecasting in smart grids. Importance of these applications is attributable to the need of accurate dynamic models for modern purposes like control, forecasting, simulation, emulation among others. Besides these two applications shown, the proposed schemes are applicable to solve very different kinds of unknown nonlinear systems with noises, uncertainties, delays, saturations, etc. Then, Chapter 7 presented the use of bio-inspired algorithms to improve neural controllers for discrete-time unknown nonlinear systems based on two approaches. First, a Neural-PSO Second-Order Sliding Mode Controller approach was presented to control a class of unknown nonlinear systems. Second, a Neural-BFO Second-Order Sliding Mode Controller for the same class of unknown nonlinear systems was incorporated. In order to show applicability of these controllers, they were applied to a Van der Pol Oscillator, and a comparative analysis was done to establish conclusions about the development of both controllers with respect to a traditional one. It was concluded that the Neural-BFO Second-Order Sliding Mode Controller had a better performance in trajectory tracking of a class of unknown discrete-time nonlinear systems with disturbances (external an internal). The results of all the simulation and experimental work done with bioinspired algorithms considered in this book demonstrate that through the use of these problem-solving approaches and tools adequate solutions may be obtained for a wide range of engineering problems; they also stand as proof of the importance and capabilities of the proposed methodologies. Finally, we emphasize the highlights of the book: • The proposed applications are very general in the sense that they are able to handle a large class of engineering problems.
132
Bio-inspired Algorithms for Engineering
Applications considered in this book include both those of simulation and experimentation. • Applications considered in this book are in the category of state-of-theart, including intelligent pattern recognition, object reconstruction, robot control and vision, and intelligent identification and control of nonlinear systems, all of which use explicit, efficient and robust systems which involve using bio-inspired algorithms to function. In regards to future work, it is possible to undertake the following: • integration of these methodologies for autonomous robotic navigation, • extrapolation of the proposed methodologies for cyberphysical systems, • expansion of the proposed methodologies to complex systems, • hybridization of bio-inspired algorithms with other soft computing techniques, • improvement of well-known bio-inspired algorithms to avoid common problems like slow convergence and stagnation in local minimal among others, and • development of new bio-inspired algorithms considering new developments for biological collective intelligence. Finally, we are certain that this book will aid to rise above limitations linked to implementing bio-inspired algorithms to solve real-life challenges and complex engineering problems that are hard to be solved using traditional methodologies. Through the new generation of students and scientists applying the ideas proposed in this book both theoretically and practically, including in their research activities, the vital importance of all the concepts contained therein will be made evident. •