AI-Powered Mobile Robot Navigation System with Prioritized Double Q-Network (PDQN) and Multi-Objective Pumafish Optimization Algorithm (MOPFOA)

EasyChair Preprint 13789

Abstract

Mobile robots are helping multiple sectors, including mining, health, space, the military, surveillance, and agriculture. Mobile robots (MR) depend mainly on complex algorithms for safe and efficient navigation. Perception, path planning, localization, and motion control are the four needs for mobile robot navigation. Although most mobile robots operate in dynamic situations, the number of algorithms able to navigate robots in such conditions is limited. This paper proposes novel reinforcement learning techniques and hybrid metaheuristic optimization for mobile robot navigation systems. Initially, the Multi-Objective Pumafish Optimization Algorithm (MOPFOA) is utilized for creating an efficient task schedule by minimizing total task completion time, path length, energy consumption, and robot idle time. After task scheduling, path planning occurs by training a deep reinforcement learning agent like a prioritized double Q-network (PDQN). This agent will plan collision-free paths, considering dynamic obstacles and optimizing multiple objectives. Additionally, novel techniques such as Multi-Agent Deep Deterministic Policy Gradient (MADDPG) are used for cooperative multi-robot navigation. After that, vision transformers (ViTs) are used for precise obstacle detection. Then the avoidance algorithm will use a hybrid of transformer-based detection and deep reinforcement learning to dynamically adjust the robot's path. Lastly, the system will guide the robot to a charging station when battery levels reach a threshold. The Python tool is used for implementing this work, and the energy consumption of the proposed work is 2.67.

Keyphrases: Deep Reinforcement Learning (DRL), Double Q-Network (DQN), Pufferfish Optimization Algorithm (POA), Puma Optimizer (PO), Vision Transformer (ViT), autonomous navigation, energy consumption, mobile robot navigation system, obstacle detection

@booklet{EasyChair:13789,
  author    = {Kuldeep Singh},
  title     = {AI-Powered Mobile Robot Navigation System with Prioritized Double Q-Network (PDQN) and Multi-Objective Pumafish Optimization Algorithm (MOPFOA)},
  howpublished = {EasyChair Preprint 13789},
  year      = {EasyChair, 2024}}