Reactive navigation under a fuzzy rules-based scheme and reinforcement learning for mobile robots.

Area of Science:

• Robotics
• Artificial Intelligence
• Machine Learning

Background:

• Mobile robot navigation requires efficient path planning and obstacle avoidance.
• Maintaining adequate battery power is critical for uninterrupted autonomous operation.
• Integrating battery management into navigation algorithms is essential for prolonged robot autonomy.

Purpose of the Study:

• To develop a machine learning algorithm for enhanced robot autonomy by optimizing battery charging decisions.
• To enable mobile robots to dynamically adjust navigation paths based on battery status and task requirements.

Main Methods:

• A hybrid machine learning approach combining reinforcement learning (RL) and fuzzy inference systems (FIS).
• The algorithm allows robots to learn optimal strategies for choosing between continuing to a destination or diverting to a charging station.
• Simulations were conducted using state representations of thirty-six and twenty states.

Main Results:

• The proposed method demonstrated faster learning convergence, requiring fewer training epochs compared to traditional RL methods.
• Robots successfully completed tasks with guaranteed battery availability across various scenarios.
• In several scenarios, the robot maintained over 80% battery charge, outperforming deterministic methods.

Conclusions:

• The integrated RL-FIS approach effectively enhances robot autonomy by intelligently managing battery power during navigation.
• This strategy ensures reliable task completion and extends operational duration for mobile robots.
• The reduced state representation (twenty states) maintained performance while improving learning efficiency.