Interception of Multiple Drone Targets by Heterogeneous Chasers using Heuristic Task Allocations with DQN-GNN Guidance Model

This paper investigates the problem of air-to-air drone interception for counter-UAS operations. The fleet of interceptors (agents) are assigned to intercept the intruders (targets) mid air at designated area. There is requirement for heuristic task allocation algorithm to assign each of the agent to track and look for specific target, for better chasing (intervention) effectiveness and better use of resources, especially in cases of heterogeneous (non-identical) chasing agents. The task allocation framework is modelled as matching and optimization problem and based on the weapon target assignment problem. The preliminary mix-integer, non-linear problem (MINLP) formulations are based on probability of interception, resource readiness and threat evaluation are used. After which, the individual agent path planning path will be done through deep Q-network with Graph Neural Network (DQN-GNN) reinforcement neural network. This learning based approach provide the guidance rules and allows the assigned agents to track the target in any environment including existing obstacles and other moving objects. Preliminary works have shown that the combination of the proposed heuristic task allocation and the Deep Q-network with Graph Neural Network (DQN-GNN) framework help more successful interception in specific time frames.