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

This paper addresses the task-priority projection problem within the Null-Space-based Behavioral Control (NSBC) framework. We propose a Reliable Intelligent Task Supervisor (RITS) for dynamic task-priority projection in networked Nonholonomic Mobile Robots (NMRs). Firstly, the NSBC task paradigm is extended to accommodate the NMRs by integrating nonholonomic constraints into both task functions and the projection operator. Subsequently, a Multi-agent Learning Task Supervisor (MLTS) is developed to allocate task priorities through environment interactions. The MLTS formulates an optimization problem for task-priority decision-making. The optimization variables are the task priorities for null-space projections, while the optimization constraints are the nonholonomic constraints. Furthermore, a Model Predictive Control Redundancy Replanner (MPCRR) is developed to enhance the reliability of the MLTS. The proposed RITS is the integration of the MLTS and the MPCRR. Simulations show that, compared to the finite state automata task supervisor, the RITS reduces average switching by approximately 36 times; compared to the model predictive control task supervisor, it reduces average iteration time by approximately 26 s. The RITS has been successfully implemented in JetAuto robots.