StanfordASL / ccscp

Chance-Constrained Sequential Convex Programming for Robust Trajectory Optimization

Chance-Constrained Sequential Convex Programming

This repository accompanies our ECC 2020 paper Chance-Constrained Sequential Convex Programming for Robust Trajectory Optimization: an algorithm for uncertainty-aware trajectory planning.

Free-flyer robot avoiding obstacles despite uncertain dynamics.

Setup

Python 3.5.2 is required. It is advised to run the following commands within a virtual environment.

	python -m venv ./venv
	source venv/bin/activate

Then, install the package as

	pip install -r requirements.txt
	pip install -e .

Demo

We provide examples for an uncertain free-flyer system and for the Astrobee robot navigating on-board the International Space Station.

	cd exps/
	python freeflyer_script.py
	python astrobee_script.py

Computation Time

We report average computation times below, measured on a laptop equipped with an Intel Core i7-6700 CPU at 2.60GHz with 8GB of RAM. As it depends on the complexity of the scenario (number of obstacles, initial/final conditions) and on the number of inner SCP iterations, we report the total time, total number of SCP iterations, and average time (= total time / nb. SCP iters).

Total Time (Nb. SCP Iters) / Avg.	N = 35	N = 30	N = 25	N = 20
Freeflyer (CoM offset, 4 obs.)	174ms (3) / 58ms	152ms (3) / 51ms	126ms (3) / 42ms	104ms (3) / 35ms
Astrobee (ISS env., 30 obs.)	3.66s (6) / 0.61s	2.40s (5) / 0.48s	1.91s (5) / 0.38s	1.53s (5) / 0.31s

Citation

@inproceedings{lew2020ccscp,
  title={Chance-Constrained Sequential Convex Programming for Robust Trajectory Optimization},
  author={Lew, Thomas and Bonalli, Riccardo and Pavone, Marco},
  booktitle={European Control Conference},
  year={2020}
}