Intelligent mobile robot navigation in unknown and complex environment using reinforcement learning technique.

Area of Science:

• Robotics
• Artificial Intelligence
• Machine Learning

Background:

• Mobile robots (MRs) are increasingly used in manufacturing, surveillance, healthcare, and warehouse automation.
• Effective control strategies are essential for safe and efficient MR operation in dynamic environments.
• Adapting MRs to unfamiliar surroundings requires advanced navigation and obstacle avoidance capabilities.

Purpose of the Study:

• To develop a new reinforcement learning (RL) technique for controlling mobile robots (MRs).
• To enhance MR navigation and collision avoidance in unknown environments using deep Q-learning (DQN).

Main Methods:

• Generated a mathematical model for MR control.
• Trained a neural network (NN) using RL, specifically a deep Q-learning (QL) agent.
• Employed the Epsilon-Greedy algorithm for simulation-based evaluation.

Main Results:

• The RL-based approach enabled autonomous learning of obstacle avoidance and navigation.
• The deep Q-Network (DQN) successfully guided MRs to goal locations in unfamiliar areas.
• The proposed method demonstrated superior efficiency and safety compared to traditional MR control strategies.

Conclusions:

• Reinforcement learning, particularly deep Q-learning, offers a powerful solution for advanced mobile robot control.
• The developed technique significantly improves MR navigation and safety in complex, unknown environments.
• This approach holds promise for enhancing automation across various industries reliant on mobile robots.