Multi-objective optimization for exploration of unknown terrain in collaborative snake robot system

Abstract

Exploration of large areas, like borders of a country, forest and areas under collapsed buildings using a team of robots has been the aspiration of the scientists since long. On this context, mapping of unknown terrain is a challenging task, especially when the target area is inaccessible by human, where mobile robot is expected to play a role in performing this task. However, this kind of mapping task is difficult to be accomplished with the traditional robots with active wheels; since they may turn over or get stuck. Moreover, using a single robot to cover a big area is not a good option, since it will take longer time to complete that task. Therefore, this study proposed a multiple snake robot system for mapping task, where the snake features of these robots allow a smooth movement on uneven surface. In this research, a mathematical model of a snake robot has been developed by modifying an existing one. Based on that model, optimal motion control parameters, namely winding angle and acceleration, were obtained using genetic algorithm multi-objective optimization technique to achieve the fastest speed of the snake robot locomotion using minimum energy consumption. On the other hand, collision among moving robots may occur when they share the same area. Therefore, a collision avoidance algorithm was developed in this research to deal with all colliding cases. In this algorithm, an imaginary rectangular safety frame is defined to be surrounding the snake robot. The benefit of this frame is to secure enough space between the robots so that any collision will be will affect to the imaginary frames not the snake robots themselves. Furthermore, multi-coverage of the same area is a common phenomenon in exploration tasks, which leads to increasing the exploration time. Therefore, two methods were proposed in this research to reduce this multi-coverage, first, by optimizing the safety frame size, secondly, by selecting the proper turning angle for collision avoidance. On the experimental side, two identical snake robots have been designed and fabricated based on simulation results. Each snake robot consists of eight links, seven active servomotors and seven free passive joints to facilitate the snake robot movement over uneven surface. One pair of passive wheels is attached to each link to generate anisotropic friction, which allows the snake robot movement along its body curve in semi-sinusoidal pattern. Both snake robots are equipped with microcontrollers, data acquisition sensors and wireless modules to communicate with a central controller. A control algorithm was developed to drive the snake robot smoothly throughout the target area including turning right or left. Simulation results showed that the objectives of minimizing energy consumption and exploration time could be achieved by decreasing the winding angle (30° was found to be the optimal). Furthermore, a trade-off between travelling speed and energy consumption was obtained using Pareto optimality, where the weighted sum method could be used to select the best solution. With respect to collaboration, the results showed that using 2.1m as the optimal safety margin besides selecting a random turning angle for collision avoidance, noticeably reduced the exploration time and thus the total energy consumption. Finally, the simulation results were verified by experiments to show the validity of the proposed system in terms of minimizing the total energy consumption and the total time needed to complete the mapping task of unknown terrain.