Learning Based Adaptive Powered Descent Guidance For Planetary Landing

Powered descent guidance (PDG) is a critical phase in planetary landing missions, where the vehicle must be steered from an initial state to a precise touchdown under stringent constraints on state and control. Classical guidance approaches—such as analytical feedback laws derived from optimal control—offer reliability and interpretability, but often rely on simplified models and fixed parameter tuning. This limits their performance in the presence of uncertainties, environmental variations, and model mismatches. 

In this work, we develop a learning-augmented adaptive guidance framework that combines the strengths of model-based optimal control with the flexibility of data-driven methods. The key idea is to retain the functional structure of classical guidance laws, ensuring stability, interpretability, and constraint satisfaction, while introducing learning layers that adapt critical parameters online or offline that are difficult to obtain analytically.

Related Publications

Quasi-Optimal Learning based Analytical Powered Descent Guidance with State and Action Constraints 

Suraj Kumar, Vaibhav Agrawal , Aditya Rallapalli, Nivriti Priyadarshini, Bharat Kumar GVP 

AIAA SCITECH 2026 FORUM, Orlando, Florida, USA

Real-Time Retargeting Using Controllability Boundary for Chandrayaan-3 Lunar Landing

Suraj Kumar, Debajyoti Chakrabarti, Aditya Rallapalli, GVP Bharat Kumar

American Control Conference 2026, New Orleans, Louisiana, USA