Dynamic Programming and Model Predictive Control Approach for Autonomous Landings

BEING lightweight and inherently easy to deploy, precision aerial delivery systems (PADSs) have the potential to lead to both drastic cost reductions and improved landing precision in the field of space exploration.With this work, we propose a technique that allows generating a PADS guidance law by using a combination of dynamic programming and model predictive control (MPC). The technique presented is based on a parallel successive-refinement dynamic programming used for the optimal trajectories generation, which is then combined with an MPC logic for optimal tracking. Although the method is more generally applicable to most celestial bodies characterized by a relatively dense atmosphere, we focus on a descent on Titan as an example to illustrate its effectiveness. Saturn moons, in fact, have been part of several research efforts due to their possible habitability conditions [1–3]. Among them, Titan has been given a lot of attention thanks to its peculiar geophysical characteristics. While it is possible to find many recent literature contributions focusing on novel guidance and control approaches for spacecraft landing on celestial bodies (e.g., [4–8]), the authors are not aware of any work that focuses specifically on parafoil landings in such environments. The different atmospheric conditions and largely unknown environment are, however, only some of the challenges that would be encountered during a similar mission. Unlike conventional powered vehicles, most parafoils do not have the ability to ascend or slow down considerably. Furthermore, given the slow entry, descent, and landing (EDL) phase due to Titan’s high atmospheric density, there is enough time for wind profiles to change considerably. Another constraint is that the final touchdown should also occur upwind to reduce the payload speed and limit the risk of a roll-over.