RotorPy

https://github.com/spencerfolk/rotorpy
Admin
A Python-based multirotor simulator with aerodynamic wrenches for education and research in UAV...
Tokens:3,605
Snippets:51
Trust Score:6.1
License:MIT
Update:4 months ago
Show doc for...
Tokens:
Raw
### Basic RotorPy Simulation Setup

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

A minimal example demonstrating how to initialize and run the RotorPy simulator with essential components like vehicle, controller, and trajectory. The initial state and simulation parameters can be configured.

```python
import numpy as np
from rotorpy.environments import Environment
from rotorpy.vehicles.multirotor import Multirotor
from rotorpy.vehicles.crazyflie_params import quad_params
from rotorpy.controllers.quadrotor_control import SE3Control
from rotorpy.trajectories.lissajous_traj import TwoDLissajous
from rotorpy.wind.default_winds import SinusoidWind

sim_instance = Environment(vehicle=Multirotor(quad_params),           # vehicle object, must be specified. 
                           controller=SE3Control(quad_params),        # controller object, must be specified. 
                           trajectory=TwoDLissajous(),                # trajectory object, must be specified. 
                           wind_profile=SinusoidWind(),               # OPTIONAL: wind profile object, if none is supplied it will choose no wind. 
                           sim_rate     = 100,                        # OPTIONAL: The update frequency of the simulator in Hz. Default is 100 Hz. 
                           imu          = None,                       # OPTIONAL: imu sensor object, if none is supplied it will choose a default IMU sensor. 
                           mocap        = None,                       # OPTIONAL: mocap sensor object, if none is supplied it will choose a default mocap.  
                           estimator    = None,                       # OPTIONAL: estimator object 
                           world        = None,                      # OPTIONAL: the world, same name as the file in rotorpy/worlds/, default (None) is empty world 
                           safety_margin= 0.25                        # OPTIONAL: defines the radius (in meters) of the sphere used for collision checking 
                       )

x0 = {'x': np.array([0,0,0]),
      'v': np.zeros(3,), 
      'q': np.array([0, 0, 0, 1]), # [i,j,k,w] 
      'w': np.zeros(3,), 
      'wind': np.array([0,0,0]),  # Since wind is handled elsewhere, this value is overwritten 
      'rotor_speeds': np.array([1788.53, 1788.53, 1788.53, 1788.53])}
sim_instance.vehicle.initial_state = x0

# The results are a dictionary containing the relevant state, input, and measurements vs time. 
results = sim_instance.run(t_final      = 20,       # The maximum duration of the environment in seconds 
                           use_mocap    = False,       # Boolean: determines if the controller should use the motion capture estimates. 
                           terminate    = False,       # Boolean: if this is true, the simulator will terminate when it reaches the last waypoint. 
                           plot            = True,     # Boolean: plots the vehicle states and commands   
                           plot_mocap      = True,     # Boolean: plots the motion capture pose and twist measurements 
                           plot_estimator  = True,     # Boolean: plots the estimator filter states and covariance diagonal elements 
                           plot_imu        = True,     # Boolean: plots the IMU measurements 
                           animate_bool    = True,     # Boolean: determines if the animation of vehicle state will play. 
                           animate_wind    = True,    # Boolean: determines if the animation will include a scaled wind vector to indicate the local wind acting on the UAV. 
                           verbose         = True,     # Boolean: will print statistics regarding the simulation. 
                           fname   = None # Filename is specified if you want to save the animation. The save location is rotorpy/data_out/. 
                    )

```

--------------------------------

### Example Cuboid with Color

Source: https://github.com/spencerfolk/rotorpy/blob/main/rotorpy/worlds/README.md

An example of a cuboid obstacle with its extents and RGB color specified. This example creates a green box.

```json
{"extents": [0, 1, -3, 2, 0, 10], "color": [0, 1, 0]}
```

--------------------------------

### Batched Simulation Example

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

Demonstrates how to use the new batched simulation environment for simulating multiple drones in parallel. This can significantly speed up simulations, especially for large numbers of drones.

```python
import rotorpy.batched_env as rpy_batched_env

# Initialize the batched environment
env = rpy_batched_env.BatchedEnv()

# Example usage (replace with actual simulation logic)
obs, info = env.reset()
for _ in range(100):
    actions = env.action_space.sample() # Sample random actions for all drones
    obs, rewards, terminated, truncated, info = env.step(actions)
    if any(terminated) or any(truncated):
        obs, info = env.reset()

env.close()
```

--------------------------------

### Install RotorPy

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

Install the base RotorPy package using pip.

```bash
pip install rotorpy
```

--------------------------------

### Reinforcement Learning Training Example (Hover)

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

An example script for training a hovering policy using reinforcement learning with the batched Gymnasium environment. It allows for faster RL training by simulating multiple vehicles in parallel.

```python
# This script is intended to be run directly to train a policy.
# See examples/ppo_hover_train.py for details.
# Note: For sim2real, further reward engineering, domain randomization,
# and hyperparameter tuning will be necessary.
```

--------------------------------

### Install RotorPy with Learning Dependencies

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

Install RotorPy along with learning-related dependencies like stable_baselines3 and tensorboard.

```bash
pip install rotorpy.[learning]
```

--------------------------------

### Benchmark Batched Simulation Speedup

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

A script to measure the speedup achieved by the batched simulation on your system. This helps in understanding the performance gains for different hardware configurations.

```python
# This script is intended to be run directly to benchmark performance.
# See examples/benchmark_batched_simulation.py for details.
```

--------------------------------

### World Bounds Definition

Source: https://github.com/spencerfolk/rotorpy/blob/main/rotorpy/worlds/README.md

Defines the spatial boundaries of the world using an 'extents' array.

```json
{
    "bounds": {"extents": [xmin, xmax, ymin, ymax, zmin, zmax]}

```

--------------------------------

### Obstacle Cuboid Definition

Source: https://github.com/spencerfolk/rotorpy/blob/main/rotorpy/worlds/README.md

Defines obstacles as cuboids with 'extents' and 'color' properties. The color is specified as RGB values between 0 and 1.

```json
    "blocks": [
        {"extents": [xmin, xmax, ymin, ymax, zmin, zmax], "color": [R, G, B]},
        ...
    ]
}

```

--------------------------------

### RotorPy Citation

Source: https://github.com/spencerfolk/rotorpy/blob/main/README.md

Use this BibTeX entry when citing RotorPy in academic publications. It includes authors, title, journal, and year.

```bibtex
@article{folk2023rotorpy,
  title={{RotorPy}: A Python-based Multirotor Simulator with Aerodynamics for Education and Research},
  author={Folk, Spencer and Paulos, James and Kumar, Vijay},
  journal={arXiv preprint arXiv:2306.04485},
  year={2023}
}
```
Tokens:
Raw