Part 1: Evaluating Multi-Unmanned Aerial Vehicles Mission Efficiency through an Agent-Based Simulation
To evaluate how well future airborne systems with autonomous components perform their missions, we need to use performance models and metrics. This study demonstrates how to determine these metrics using a simulation that involves multiple Unmanned Aerial Vehicles (UAVs) as agents in a 2D environment. The simulation focuses on an air-to-ground operation with multiple UAVs assigned with different tasks. By using different types of UAV models with different characteristics in the simulation, we can compare their performances and determine which ones are more efficient and suitable for the mission.

Part 2: Interception of Multiple Drone Targets by Heterogeneous Interceptors using Heuristic Task Allocations with Graph-based Neural Network

• To optimise the effectiveness of air-to-air drone interception during counter-Unmanned Aerial Systems (UAS) operations, we employ a heuristic task allocation algorithm for a scenario where different types of interceptors (agents) are tasked with intercepting intruder drones (targets) at specific locations in the sky. This algorithm assigns each interceptor to track and search for a specific target, based on the concept of weapon target assignment.
• To guide the assigned interceptors in tracking the targets, we use a graph-based neural network approach. This approach provides rules for navigation and allows the interceptors to track the target in any environment, even when there are obstacles or other moving objects present. Preliminary studies have shown that the proposed heuristic task allocation and graph-based neural network significantly improve the efficiency of drone interception.