International Journal of Engineering and Technology

An Adaptive Dynamic Programming and Reinforcement Learning (ADPRL) based instant path planning algorithm is proposed in this paper. The layout of any indoor environment is always known. This information is converted into a binary matrix containing free space and obstacle space using image processing system. A dynamic program algorithm translates the rough obstacles to expected shaped obstacles so the robot is not confined in motion. A grid policy is used for value evaluation of reward function. Value iteration draws out all possible paths from goal to target. A Q-learning algorithm finds the best possible path from the numerous possible paths determined. A Biezer curve based approximation is done to smoothen the discrete way points for smooth motion and determination of linear and angular velocities for a differential drive robot. The simulation and the results show the proposed algorithm have better processing time, less computational complexity, and instant determination of path, compared to other existing methods.