Optimal task-priority projection for behavioral control of networked nonholonomic Mobile robots.

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• The Null-Space-based Behavioral Control (NSBC) framework requires effective task-priority projection.
• Networked Nonholonomic Mobile Robots (NMRs) present unique challenges due to their kinematic constraints.

Purpose of the Study:

• To develop a dynamic task-priority projection method for NMRs within the NSBC framework.
• To introduce a Reliable Intelligent Task Supervisor (RITS) for enhanced control performance.

Main Methods:

• Extended NSBC to incorporate nonholonomic constraints into task functions and projection operators.
• Developed a Multi-agent Learning Task Supervisor (MLTS) for adaptive task-priority allocation.
• Integrated MLTS with a Model Predictive Control Redundancy Replanner (MPCRR) to form RITS.

Main Results:

• RITS demonstrated a significant reduction in average switching (approx. 36x) compared to finite state automata supervisors.
• RITS achieved a substantial decrease in average iteration time (approx. 26s) versus model predictive control supervisors.
• Successful implementation and validation of RITS on JetAuto robots.

Conclusions:

• The proposed RITS effectively addresses dynamic task-priority projection for NMRs.
• RITS offers superior performance in terms of switching frequency and iteration time.
• The RITS framework is practical and has been successfully deployed on real-world robotic platforms.