Lunar Lander Optimal controlled descent using Model Predictive Control

I'm working on a project focused on developing a guidance and control system for a lunar lander using Model Predictive Control (MPC). The goal is to ensure a smooth and efficient landing by optimizing the lander's descent trajectory while minimizing fuel consumption. The system accounts for mass depletion, thrust vectoring, rotational motion, and external disturbances.

The lander’s movement is modeled in six degrees of freedom (6-DOF), and an MPC-based controller computes the best thrust and orientation adjustments at each timestep. To validate performance, we have a high-fidelity simulation implemented in Python using CasADi for nonlinear optimization. The system considers trajectory constraints, actuator limits, and disturbances while maintaining stability. The simulation helps analyze state trajectories, velocity changes, angular motion, and fuel usage to ensure the controller performs as expected.

I’ve made significant progress on the open-loop dynamics, which solve the optimal control problem and generate control inputs. However, implementing a closed-loop solution has been challenging. While the open-loop works perfectly, the closed-loop version struggles to find solutions efficiently after about 20 horizon steps—either failing to converge or taking excessive iterations.

I have two files:

The first file solves the optimal control problem and generates control inputs.
The second file applies these control inputs to real-world dynamics while accounting for disturbances.
Would you be available to review my MPC_02 file and help troubleshoot the issue? If you think you can assist in refining the controller, optimizing computational efficiency, and improving real-time feasibility, I’d be happy to discuss further and schedule an online meeting at a convenient time.

Looking forward to your thoughts!