Genetic Fuzzy Systems for Decentralized, Multi-UAV Cargo Handling
Caleb Bisig, Jorge B. Montejo, Matthew R. Verbryke, Anoop Sathyan, Ou Ma · AIAA Scitech 2020 Forum · 2020
Genetic-fuzzy control methodology is utilized to enable an intelligent and adaptive multi-UAV system to transport a load attached with cables to a given target location. The UAVs (considered as simplified quadcopters) are controlled in a decentralized fashion, reducing communication complexity, and increasing the adaptability and success rate of the overall system. A set of Fuzzy Inference Systems (FIS) tuned via a Genetic Algorithm determines the behavior of each quadcopter based on its current state, the state of the load, and its relation to a set of its neighbors. The combination of all the drones’ behaviors results in group intelligence and the completion of the assigned task. The controllers are created and tuned using a cascading Genetic Fuzzy Tree structure to minimize computational cost. Results show the collective of drones transporting the load progressively closer to a target location over 100 generations, while avoiding collisions with one another and the payload. The proposed architecture shows the potential for intelligent controllers enabling adaptable and collaborative systems.