Human-Centered UAV–MAV Teaming in Adversarial Scenarios via Target-Aware Intention Prediction and Reinforcement Learning

Abstract

Tacit understanding refers to the ability of team members to work together seamlessly and intuitively without explicitly communicating in detail. This ability is crucial for effective teamwork in complex situations that involve both manned and unmanned aerial vehicles (UAVs). Existing collaborative tasks between manned and unmanned aircraft focus mainly on optimizing communication and the UAVs’ flight paths but neglect the benefits of tacit and intelligent operational cooperation with pilots. To address this limitation, we propose a tacit collaborative attack method that utilizes the UAVs’ capacity for tacit understanding to infer human intent and select the appropriate targets for collaborative attack missions. A learning framework incorporating intention prediction and reinforcement learning paradigms is developed to teach the UAV to generate corresponding collaborative attack actions. Finally, we present results from extensive simulation experiments in a homemade game environment to demonstrate the efficiency and scalability of our method within the proposed framework. The video can be found at https://www.youtube.com/watch?v=CjXhkD7ko14.