Trajectory optimization of wheeled mobile robot (WMR) in the wall-type arena

Abstract

Over the last few decades, navigational motion synthesis in cluttered arena has remained as one of the most challenging area of investigation. In this paper, a smart neural network (SNN) based navigation control architecture for a wheeled mobile robot (WMR) has been presented. Here, two responsive activities are considered as the input function of the SNN controller. The obstacle position and obstacle angle with reference to the target location are two activities followed by the WMR during the trajectory planning in an environment. The outputs for the controller are wheel velocity and steering angle followed by target direction. To minimize the errors of the weight function, the backpropagation training architecture has been analyzed. It has been observed that, the robot generates a quick reaction time by using the SNN methodologies as compared to other methodologies. To evaluate the effectiveness of the developed controller, the proposed architecture has been tested in both simulation and experimental platforms keeping common environmental conditions such as position of source, target and obstacles. The results obtained from both of the platforms have been compared in terms of selected navigational parameters and the deviation obtained is found to be within 8%. The V-Rep simulation software has been used for the simulation experiments and Khepera 3 mobile robot has been used for real time experiments.