A joint opinion dynamics and performance-driven approach for state estimation fusion of multi-agent systems

This study develops a distributed adaptive sensor fusion architecture to address state estimation problems for multi-agent systems made up of two distinct node classes: sensors and agents. Evaluating their reliability specifically addresses the issue of suitably combining the sensor data that is sent to the agents and originates from the agents on the basis of their trustworthiness. This is accomplished by fusing concepts from the theory of consensus of multi-agent system configurations with the so-called Perturb&Observe approach well established in power electronics. This results in the creation of a novel performance-driven adaptive algorithm that can choose the more suitable and reliable sensor groups responsible for enhancing agent state estimates based on a specified performance criterion. The final test of the proposed methodology is conducted on a benchmark simulation case involving a group of drones in path-planning tasks under obstacle avoidance prescriptions.