Browse Teach Curate

Certifiably Stable Learning-Based Optimal Guidance for Spacecraft Rendezvous via Control Lyapunov Functions

A guidance framework that couples a positive-definite neural control Lyapunov function with supervised learning on optimal state-control pairs, defining a thrust policy that minimizes the Lyapunov derivative to jointly certify stability and produce near-optimal, even bang-bang fuel-optimal, guidance for Clohessy-Wiltshire rendezvous dynamics and general control-affine systems.

Data Text

Learning-Based Stable Optimal Guidance for Spacecraft Close-Proximity Operations Kun Wang∗1

Spacecraft rendezvous requires a chaser to match a target's position and velocity, and doing so with minimum time or fuel is naturally posed as an optimal control problem. Machine-learning guidance c…