Fast Prediction of Reachable Area for Reusable Vehicle under Waypoint Constrains using a Multiobjective DRL Heuristic Algorithm

A reachable area defines the potential end positions that a reusable vehicle can reach under its current flight state. The area comprises feasible flight trajectories and helps to sustainably regenerate trajectories through instantaneous computation. Thus, it is essential for online plan decisions, particularly in emergencies. However, computing end positions can be challenging when there is high degree of nonlinear coupling between the flight constraints, dynamics, objective functions, and thermal / load factors. To solve this issue, we propose a new flight profile that easily satisfies strict constraints. In addition, we devise a series of analytic dynamics to instantly predict the thermal / load factors and objective functions. The analytic models integrate Earth's perturbations and rotations and complex aerodynamic models, thus improving the accuracy of the derived reachable area. Based on a novel Mercator plane with minimum projection deformation, we devise a lateral profile using a cubic B-spline model that easily satisfies the extensively checked waypoints. Based on the results, the proposed algorithm mitigates nonlinear coupling to a significant extent and facilitates reachable area generation. In addition, we study and reveal the principles of evolutionary method improvements and convergence and devise a deep reinforcement learning-based multiobjective particle swarm optimization method to obtain the Pareto Set. The Set maximizes / minimizes the down/crossrange of a vehicle and comprises a series of end positions to define the reachable area. The reachable area is then employed to rapid vehicle trajectory recomputation. Finally, we validate our algorithm through numerical simulations analyze the results for future improvements