Awesome model_predictive_control created by AdityaNG
Project description
model_predictive_control
Python implementation of MPC solver
Install it from PyPI
pip install model_predictive_control
Usage
import numpy as np
from model_predictive_control.cost.trajectory2d_steering_penalty import (
Traj2DSteeringPenalty,
)
from model_predictive_control.models.bicycle import (
BicycleModel,
BicycleModelParams,
)
from model_predictive_control.mpc import MPC
# Initialize the Bicycle Model
params = BicycleModelParams(
time_step=time_step,
steering_ratio=13.27,
wheel_base=2.83972,
speed_kp=1.0,
speed_ki=0.1,
speed_kd=0.05,
throttle_min=-1.0,
throttle_max=1.0,
throttle_gain=5.0, # Max throttle corresponds to 5m/s^2
)
bicycle_model = BicycleModel(params)
# Define the cost function
cost = Traj2DSteeringPenalty(model=bicycle_model)
# Initialize MPC Controller
horizon = 20
state_dim = 4 # (x, y, theta, velocity)
controls_dim = 2 # (steering_angle, velocity)
mpc = MPC(
model=bicycle_model,
cost=cost,
horizon=horizon,
state_dim=state_dim,
controls_dim=controls_dim,
)
# Define initial state (x, y, theta, velocity)
start_state = [0.0, 0.0, 0.0, 1.0]
# Define desired trajectory: moving in a straight line
desired_state_sequence = [[i * 1.0, i * 0.5, 0.0, 1.0] for i in range(horizon)]
# Initial control sequence: assuming zero steering and constant speed
initial_control_sequence = [[0.0, 1.0] for _ in range(horizon)]
# Define control bounds: steering_angle between -0.5 and 0.5 radians,
# velocity between 0.0 and 2.0 m/s
bounds = [[(-np.deg2rad(400), np.deg2rad(400)), (-1.0, 1.0)] for _ in range(horizon)]
# Optimize control inputs using MPC
optimized_control_sequence = mpc.step(
start_state_tuple=start_state,
desired_state_sequence=desired_state_sequence,
initial_control_sequence=initial_control_sequence,
bounds=bounds,
max_iters=50,
)
Run the demo with the following
$ python -m model_predictive_control
#or
$ model_predictive_control
Cite
This work was a part of the D³Nav paper. Cite our work if you find it useful
@article{NG2024D3Nav,
title={D³Nav: Data-Driven Driving Agents for Autonomous Vehicles in Unstructured Traffic},
author={Aditya NG and Gowri Srinivas},
journal={The 35th British Machine Vision Conference (BMVC)},
year={2024},
url={https://bmvc2024.org/}
}
Development
Read the CONTRIBUTING.md file.
TODO
- Bicycle Model
- Drone Model
- MPC
- Visualizer Demo
- MPC Auto-Optimizer: Takes a set of expected vehicle trajectories and the search space of hyperparamters and returns the list of optimal hyperparameters
- MPC Compiler: Takes the MPC model with a set of expected vehicle trajectories and produces numpy array a mapping from trajectory to control signals. This can be used with a cosine similarity logic to decide on control logic in real time.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file model_predictive_control-0.4.0.tar.gz
.
File metadata
- Download URL: model_predictive_control-0.4.0.tar.gz
- Upload date:
- Size: 18.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.12.5
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 5d812faf790f7df54c4308914f30d4757809d6de0a131712fc817942508a8236 |
|
MD5 | 094382a26c0d329cd545a1cbc3e1aa3e |
|
BLAKE2b-256 | 2dd7bc2ff7874919a90b709a09d4e15d6edbab6582878d8c8517e79dda37efc5 |
Provenance
File details
Details for the file model_predictive_control-0.4.0-py3-none-any.whl
.
File metadata
- Download URL: model_predictive_control-0.4.0-py3-none-any.whl
- Upload date:
- Size: 18.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.12.5
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | e4b5514684a2ea2a1270e7617b0f2cfa12aa63b3041ede965110c6bd968259b0 |
|
MD5 | 90d15777696aac06801226e6e5a8b1f7 |
|
BLAKE2b-256 | caf50057d3708f50299551bff0e211100e3ebf3448f4f9d766cf0c70b196d26f |