Overview
Designing spacecraft trajectories often boils down to solving a computationally expensive optimal control problem. Spacecraft often have limited onboard computing capabilities, which requires them to rely on ground-based trajectory design. This work is focused on using computationally lightweight Neural Networks to approximate the solution to an optimal control problem in real-time and opens up exciting possibilities in the realm of spacecraft autonomy.
Papers (Reverse Chronological Order)
Measuring Resilience Of Autonomous Controllers To Missed Thrust Events On A Mars To Earth Trajectory -ESA Conference on Guidance Navigation and Control – With Rohan Sood & Frank Laipert – We setout different metrics and methods for determining how effective different autonomous controllers are in autonomously repairing a trajectory after a missed thrust event
Using Reinforcement Learning to Design Missed Thrust Resilient Trajectories – Astrodynamics Specialist Conference 2020 – With Kyra Bryan, Rohan Sood & Frank Laipert – We explore how reinforcement learning can be used to improve a neural network’s recovery from missed thrust events.
Neural Network Optimal Control in Astrodynamics: Application to the Missed Thrust Problem – Acta Astronautica – With Rohan Sood & Frank Laipert – We explore how a neural network can be used to autonomously navigate a spacecraft and even recover from missed thrust events in a range of astrodynamic environments.
Neural Network Based Optimal Control: Resilience to Missed Thrust Events for Long Duration Transfers– Astrodynamics Specialist Conference 2019 – With Rohan Sood & Frank Laipert – We explore how a computationally lightweight neural network can be used to autonomously recover from missed thrust events.