Deep Reinforcement Learning for Dynamic Obstacle Avoidance of Mobile Robots in Indoor Environments: A Review

Area of Science:

• Robotics
• Artificial Intelligence
• Machine Learning

Background:

• Autonomous navigation in indoor environments requires effective dynamic obstacle avoidance.
• Traditional algorithms struggle with unstructured and unpredictable indoor scenes, limiting adaptability and flexibility.
• Deep Reinforcement Learning (DRL) offers a promising solution due to its decision-making, learning, and complex system modeling capabilities.

Purpose of the Study:

• To provide a comprehensive review of Deep Reinforcement Learning (DRL) for indoor dynamic obstacle avoidance.
• To categorize fundamental DRL algorithms and discuss their applications.
• To identify challenges and outline future research directions in this domain.

Main Methods:

• The paper reviews the theoretical foundations of DRL.
• It categorizes DRL algorithms into Value function-based, Policy-based, and Actor-Critic-based approaches.
• It summarizes improvement strategies for basic algorithms and analyzes their performance impacts.

Main Results:

• DRL algorithms show significant potential for improving dynamic obstacle avoidance in mobile robots.
• Various DRL approaches offer different trade-offs in adaptability and flexibility for indoor environments.
• Identified core challenges and effective improvement strategies for DRL in this context.

Conclusions:

• DRL is a key technology for advancing autonomous navigation and obstacle avoidance in mobile robots.
• This review serves as a systematic reference for DRL algorithm design and engineering applications.
• Future research should focus on addressing remaining challenges and exploring new DRL advancements for enhanced robot autonomy.