6-881 Final Project - Motion Planning

This repository hosts the final project for MIT 6.881 Intelligent Robot Manipulation offered for 2018 Fall (09/05/2018 - 12/12/2018). The project uses several tools accompanying the course.

Project Description

This project implements and evaluates three robotic motion planning algorithms:

• Rapidly-exploring Random Tree (RRT)
• Rapidly-exploring Random Tree Star (RRT*)
• Lazy Shortest Path (LazySP)

The algorithms are developed in the context of the KUKA iiwa arm, using the Drake toolbox.

Pre-reqs

In the root directory of this repository, run the following command in a terminal to build a docker image that includes Drake:

$ docker build -t motion-planning -f ubuntu16_04_mit6881.dockerfile --build-arg DRAKE_VERSION=20181203 .

Use

In the root directory of this repository, run

$ python ./docker_run.py --os [your operating system]

where [your operating system] should be replaced with mac or linux. This command will start a docker container (virtual machine) with the docker image you have created. The Motion-Planning folder on the host machine (your laptop/desktop) is mounted to /Motion-Planning in the docker container.

In the docker container, run

$ terminator

to launch terminator, a popular terminal multiplexer on linux. The terminator window is launched with a dark green background to distinct itself from terminals running on the host machine.

Run Motion Planning

$ python motion_planner/run_reach_brick

Its arguments include:

• -a {algorithm} where {algorithm} could be RRT, RRTStar, and LazySP. Default is LazySP.
• -m {max_iter} where {max_iter} is the maximum number of iterations to run. Default is 300.

Video Demo

Video Demos of the three algorithms deployed on real KUKA iiwa arm can be found at:

• RRT
• RRT*
• LazySP