Autonomous Robot Navigation Based on Depth Deterministic Policy Gradient

Abstract

Generally speaking, robot navigation can take the place of people in hazardous work contexts including military settings and fires. Furthermore, these robots are frequently assigned to carry out repetitive motions such as delivering things inside like places and the limited observable environment of autonomous robots causes the training efficiency of route planning method to be poor and convergence speed to be slow and is applied to navigation robot path planning. In this manuscript, Autonomous Robot Navigation Based on Depth Deterministic Policy Gradient (DDPG-FFOA-ARN-RL) is proposed. Initially input data are gathered from Indoor Robot Navigation Dataset (IRND). To execute this, input data is pre-processed using Switching Hierarchical Gaussian Filter (SHGF) and it removes the noise from collected data; then the Pre-processed data are fed to Depth Deterministic Policy Gradient (DDPG) for effectively categorize the autonomous robot navigation. Then using the Reinforcement Learning the robot navigation is identified. Generally, DDPG doesn’t express adapting optimization approaches to determine optimal parameters to ensure accurate autonomous robot navigation based depth deterministic policy gradient. Hence, the Fennec Fox Optimization Algorithm (FFOA) to optimize Depth Deterministic Policy Gradient; which accurately navigates the autonomous robot navigation. Then the proposed DDPG-FFOA-ARN-RL is implemented and the performance metrics like Success Rate, Navigation Time, Path Efficiency, Collision Rate and Energy Consumption around are analyzed. Performance of the DDPG-FFOA-ARN approach attains 18.75%, 26.89% and 32.57% higher Success Rate; 17.02%, 23.26% and 32.42% lower Navigation Time and 18.43%, 25.64% and 31.40% lower Energy Consumption when analyzed through existing techniques like Deep Deterministic Policy Gradient-depend Autonomous Driving for Mobile Robots in Sparse Reward Environments (DDPG-ADMR-SRE), automatic driving control technique depend on deep deterministic policy gradient (ADC-DDPG), multilayer decision-based fuzzy logic method to navigate mobile robot in unknown dynamic environments (MD-FLM-NMR) methods respectively.