Tactical Conflict Solver Assisting Air Traffic Controllers Using Deep Reinforcement Learning

To assist air traffic controllers (ATCOs) in resolving tactical conflicts, this paper proposes a conflict detection and resolution mechanism for handling continuous flow by adopting finite discrete actions to resolve conflicts. The solver (TCS) was developed based on deep reinforcement learning (DRL) train TCS agent with the actor–critic using Kronecker-factored trust region. agent’s are determined ATCOs’ instructions, such as altitude, speed, heading adjustments. reward function is designed accordance control regulations. Considering uncertainty real-life situation, study characterised deviation of aircraft’s estimated position improve feasibility schemes. A DRL environment actual airspace structure density operation simulation system. Results show that 1000 test samples, trained could 87.1% samples. rate decreased slightly 81.2% when increased factor 1.4. This research can be applied intelligent decision-making systems control.