Closed loop Planning and Control of Low Thrust Orbit Transfers Using Physics Informed Evolutionary Machine Learning

This project develops closed-loop planning and control methods for long-duration, low-thrust orbital transfers, with a focus on real-time onboard implementation for electric propulsion missions. Classical feedback approaches such as Q-law are widely used for multi-revolution orbit transfers, but their lack of formal stability guarantees limits robustness and operational reliability. To address this, the project introduces a Lyapunov-guided modification of Q-law that ensures closed-loop stability while retaining computational efficiency suitable for onboard execution. Building on this stable control structure, the framework integrates physics-informed evolutionary machine learning to adaptively optimize control gains for performance objectives such as minimum transfer time or fuel consumption. Control gains are parameterized as state- and time-dependent functions and optimized using a hybrid simulation-based evolutionary strategy, combining structured random sampling with covariance matrix adaptation. The resulting learning-augmented controller preserves stability and convergence guarantees. For flight readiness, the learned control policies are distilled into low-complexity polynomial representations for robust onboard deployment.

Related Publications

Adaptive Q-law Control for Closed-loop Electric Propulsion Orbit Transfer

Suraj Kumar, Sri Aditya Deevi, Aditya Rallapalli, GVP Bharat Kumar 

AIAA SciTech Form 2026, Orlando, Florida, USA  

Closed-Loop Control Law for Low Thrust Orbit Transfer with Guaranteed Stability  

Suraj Kumar, Aditya Rallapalli, Bharat Kumar GVP 

Indian Control Conference (ICC) 2025, Bengaluru, India