Naive Model Predictive Control (MPC) Implementation

This repository contains a naive implementation of Model Predictive Control (MPC) that uses linearized dynamic models for path prediction and control.

Implementation Overview

This is a simplified MPC approach that:

1. Linearizes the dynamic model around the current state
2. Uses this linearized model to predict the system's behavior over the next N time steps
3. Optimizes control inputs without a reference trajectory, instead using a target point in the loss function

Key characteristics:

• No reference trajectory is used - optimization targets a single point
• The linearized dynamics are computed once at each time step based on the drone's current (real) state and are then used to simulate the entire predicted (future) trajectory.
• Simple quadratic cost function penalizing distance to target and control effort

Setup Instructions

pip install -r requirements.txt

Usage

python vis.py