Abstract:
Physics-informed deep reinforcement learning (RL) is proposed for ATM aircraft conflict resolution. To enable optimal policy searching and human-explainable results for display and decision-making, the learning algorithm integrates prior physics understanding and model. First, the solution space diagram (SSD) used in the ATM for conflict detection and mitigation integrates intruder quantities, speeds, heading angles, and positions into an image. The SSD provides prior ATM physics knowledge for learning. Deep reinforcement learning uses SSD images and convolution neural networks. Next, an actor-critic network learns conflict resolution policy. Numerical examples demonstrate the methodology. Physics-informed learning explores discrete and continuous RL. The algorithm is compared to classical RL-based conflict resolution. The encoded prior physics understanding allows the proposed approach to handle any number of intruders and converge faster. The learned optimal policy also helps display decision-making. The current investigation yields several major conclusions and future work.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
