🚁 mudrone

Quadrotor (Crazyflie 2) trajectory tracking in MuJoCo physics simulation, implementing two Model Predictive Control approaches: MPC (CasADi nonlinear optimization) and MPPI (sampling-based Path Integral control).

MPC — Circle	MPC — Figure-8

---

📁 Project Structure

mobile_mujoco/
├── models/
│   └── drone/
│       └── bitcraze_crazyflie_2/
│           ├── cf2.xml          # Crazyflie 2 MJCF model (with spinning props)
│           ├── scene.xml        # Simulation scene (ground + lighting)
│           └── assets/          # 3D mesh files
├── drone_mpc/
│   ├── drone_env.py             # MuJoCo environment wrapper
│   ├── mpc_controller.py        # MPC controller (CasADi + IPOPT)
│   ├── mppi_controller.py       # MPPI controller (sampling-based)
│   ├── inner_loop.py            # Attitude PD inner loop (200 Hz)
│   ├── trajectory.py            # Trajectory generators
│   └── visualization.py        # Plotting utilities
├── run_mpc.py                   # Run MPC tracking
├── run_mppi.py                  # Run MPPI tracking
├── run_compare.py               # MPC vs MPPI comparison
└── environment.yml              # Conda environment

---

🔧 Installation

1. Create Conda environment

conda create -n mobile_mujoco python=3.10 -y
conda activate mobile_mujoco

2. Install dependencies

conda env create -f environment.yml
conda activate mobile_mujoco

---

🚀 Usage

▶️ MPC Trajectory Tracking

# Basic run (no viewer, generates result plots)
python run_mpc.py

# With real-time MuJoCo viewer
python run_mpc.py --render

# Custom parameters
python run_mpc.py --radius 1.5 --height 1.2 --omega 0.3 --duration 30 --render

# Figure-8 trajectory
python run_mpc.py --trajectory lemniscate --radius 2.0

# Save result plot
python run_mpc.py --save results/mpc_circle.png

▶️ MPPI Trajectory Tracking

# Basic run
python run_mppi.py

# With viewer
python run_mppi.py --render

# Tune MPPI parameters
python run_mppi.py --n-samples 512 --temperature 0.02 --horizon 40

# Save result
python run_mppi.py --save results/mppi_circle.png

⚖️ MPC vs MPPI Comparison

python run_compare.py
python run_compare.py --radius 1.0 --omega 0.5 --duration 30
python run_compare.py --save results/comparison.png

🎛️ CLI Arguments

Argument	Default	Description
--radius	1.0	Trajectory radius (m)
--height	1.0	Flight altitude (m)
--omega	0.5	Angular speed (rad/s)
--duration	20.0	Simulation duration (s)
--trajectory	circle	circle | lemniscate
--render	False	Enable MuJoCo real-time viewer
--save	None	Save plot to file path
--horizon	25/30	MPC/MPPI prediction horizon steps
--n-samples	256	MPPI sample count K
--temperature	0.05	MPPI temperature λ (lower = greedier)

---

📝 TODO

• Add warmup and smoothing fixes for MPC/MPPI startup.
• Improve MPPI performance by optionally integrating JAX (vectorized sampling / GPU).
• Add f1tenth env.