[New Feature] Model Predictive Control(MPC) for Path Tracking

[mohitk3000]

Description

The controller uses a receding-horizon NLP formulation built on CasADi and do-mpc, with a discrete-time bicycle kinematic model, time-varying reference trajectory (TVP), and IPOPT as the backend solver. The public interface mirrors MppiController, so either controller can be swapped into the same simulation loop without changes.

---

Checkpoints:

• Cost Function — Implemented stage cost (tracking error + control effort) and terminal cost (heavier tracking weights), with angle-safe yaw error via atan2.
• Smoothness Penalty — Add rterm (Δu penalty) via mpc.set_rterm() to penalize rapid control changes
• Hard Constraints — Enforce box constraints on steering angle, acceleration, and vehicle speed
• Solver Integration — Install CasADi and do-mpc in the Docker container.
• Sample Simulation — Created a simulation script to follow a reference course.
• Visualization — Plot the MPC predicted (optimal) trajectory as a dashed line alongside the vehicle.