Constraint-Informed Learning for Warm-Starting Trajectory Optimization

摘要

Future spacecraft and surface robotic missions require increasingly capable autonomous systems for exploring challenging and unstructured domains, and trajectory optimization will be a cornerstone of such autonomous operations. However, the nonlinear optimization solvers required remain too slow for use on relatively resource-constrained flight-grade computers. In this work, we turn toward amortized optimization, a learning-based technique for accelerating optimization run times, and develop Trajectory Optimization with Merit Function Warm Starts. Offline, using data collected from a simulation, we train a neural network to learn a mapping to the full primal and dual solutions given the problem parameters. Crucially, we build upon recent results from decision-focused learning and develop a set of decision-focused loss functions using the notion of merit functions for optimization problems. We show that training networks with such constraint-informed losses can better encode the structure of the trajectory optimization problem and jointly learn to reconstruct the primal-dual solution while yielding improved constraint satisfaction. Through numerical experiments on a lunar rover problem and a 3-degree-of-freedom Mars powered descent guidance problem, we demonstrate that Trajectory Optimization with Merit Function Warm Starts outperforms benchmark approaches in terms of both computation times and network prediction constraint satisfaction.