Autonomous mobile robot navigation in uncertain dynamic environments based on deep reinforcement learning

RL-Based Sim2Real Enhancements for Autonomous Beach-Cleaning Agents

“…However, recent advancements in DRL and simulation technologies have introduced new opportunities for improving these systems. Studies employing DRL agents, such as DDPG and DQN, have been conducted solely in simulation [16], demonstrating DRL's capability to train agents that can navigate and operate in complex [17,18] and dynamic [6,19] environments, typical of coastal settings. The success of DRL in autonomous vehicles [20,21] and drones [22][23][24] highlights its potential for enhancing autonomous navigation and real-time decision-making, which are crucial for efficient beach cleaning.…”

“…Building upon existing knowledge in the fields of robotics and artificial intelligence, this paper employs DRL, an advanced AI technique that integrates deep learning with reinforcement learning principles, to develop effective control systems for autonomous robots [4,5]. Through extensive training in simulated environments, these robots master essential skills for navigating complex terrains, avoiding obstacles, and maintaining precise position control, crucial for their effective operation in real-world scenarios [6].…”