Online and Robust Intermittent Motion Planning in Dynamic and Changing Environments

Area of Science:

• Robotics
• Artificial Intelligence
• Control Theory

Background:

• Dynamic environments pose challenges for robot motion planning.
• Unknown robot dynamics and external disturbances complicate control.
• Online and intermittent planning is crucial for adaptability.

Purpose of the Study:

• To propose RRT-, an online and intermittent kinodynamic motion planning framework.
• To address challenges of unknown robot dynamics and disturbances.
• To ensure closed-loop stability in dynamic environments.

Main Methods:

• Leveraging Rapidly-exploring Random Trees (RRT) for global path planning and replanning.
• Formulating waypoint generation as a sequence of boundary-value problems (BVPs).
• Employing a robust intermittent Q-learning controller with a relaxed persistence of excitation technique.

Main Results:

• The proposed framework effectively plans motion in dynamic environments.
• The Q-learning controller converges to optimal control despite unknown dynamics.
• Lyapunov-based proofs confirm closed-loop stability.

Conclusions:

• RRT- provides a robust solution for kinodynamic motion planning in uncertain dynamic environments.
• The integration of RRT and Q-learning offers adaptive and stable robot control.
• Numerical experiments validate the framework's effectiveness.