DECENTRALIZED TRAJECTORY PLANNING AND EVENT-TRIGGERED COMMUNICATION FOR MULTI-UAV NAVIGATION

Abstract

This thesis provides a decentralized trajectory planning framework to control multiple Unmanned Aerial Vehicles (UAVs) in complex three-dimensional space. It integrates sampling-based path planning with Rapidly-exploring Random Tree (RRT) and event-triggered communication with Virtual Potential Fields (VPF) control mechanism to offer collision and deadlock-free navigation. The event-triggered mechanism reduces communication overhead by transmitting information as needed. In addition to this, a token-based prioritization mechanism dynamically manages UAV interactions and calculates repulsive force strength to offer efficient coordination in dense obstacle space.

Simulation experiments tested performance with constrained paths and scenarios with bottlenecks, analyzing collision avoidance, deadlock handling, and communication efficiency at different UAV densities. Experiments proved that the framework supported maintaining collision-free paths, effective deadlock resolution, and less communication overhead compared to centralized approaches.

Practical applications must all be supplemented by strong state estimation methods, including sensor fusion, for coping with real-world uncertainties. Later studies are suggested to consider dynamic barriers and delays in communications for improving real-world applicability.