Skip to content

Latest commit

 

History

History
131 lines (93 loc) · 7.12 KB

README.md

File metadata and controls

131 lines (93 loc) · 7.12 KB
Raw

TRACE

© 2023 Massachusetts Institute of Technology

  • Latest Known Flight Software Compatbility: v16.1
  • Stable branch: main

TRACE (Tumbling Rendezvous via Autonmous Characterization and Execution) is a set of software modules that enable autonomous rendezvous with arbitrary tumbling targets. A significant percentage of space debris like rocket bodies and defunct satellites are in uncontrolled tumbles, often with unknown tumble characteristics. In order to soft dock for servicing, refueling, or de-orbit, a servicer spacecaft must figure out the nature of the target's tumble and safely reach the target. This often eliminates teleoperation as a feasible option---this is where TRACE comes in. TRACE has been demonstrated on-orbit over multiple test sessions on the International Space Station using NASA's Astrobee robots.

Background

TRACE was developed as part of the ROAM/TumbleDock project, a collaboration between the MIT Space Systems Lab and DLR's Institute of Robotics and Mechatronics. TRACE observes a tumbling Target, estimates its state/parameters, creates a motion plan to reach it, and then plans/tracks robustly in real-time to reach a safe rendezvous position known as the "mating point." This code has flown multiple times on the International Space Station, verifying the real-time performance of the various algorithmic components on hardware. The work, including its assumptions and experimental results, is summarized in detail in few useful references below:

This repository includes modules for target time-of-flight camera-based state estimation, robust model predictive control, ROS-based coordination code, and sample trajectories. It does not contain DLR's separate motion planning method---a suitable motion planner must be substituted.

This code can also interface with Astrobee's simulation environment for convenient testing. A brief guide has also been produced for Astrobee's Flight Software to aid in any Astrobee-related development.

Code Structure

TRACE is designed to run on a debian-based system, and has recently been used with both Ubuntu 16.04 and 20.04. Your base system should be a bare version of either 16.04 or 20.04 (recommended).

Note: You might also choose to start from NASA Astrobee's preconfigured Ubuntu 20.04 container if you prefer working with Docker. See below.

TRACE's modules are general-purpose. In this repo, they are integrated with ROS for communication. Full simulation tests (and Astrobee hardware integration) of TRACE are possible using NASA's Astrobee simulator. Discussion of this setup is described below.

Dependencies and Setup

TRACE's major dependencies include:

  • GTSAM
  • CasADi
  • Teaser++
  • ROS (for messaging, Astrobee interfacing, and tf convenience functions)

If you haven't configured ROS yet, add its package lists:

sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'
sudo apt install curl # if you haven't already installed curl
curl -s https://raw.githubusercontent.com/ros/rosdistro/master/ros.asc | sudo apt-key add -
sudo apt update

Most of TRACE's dependencies are satisfied on a debian-based system with:

# ROS and core dependencies
sudo apt install pip3 python-is-python3 python3-rospkg-modules libgmp3-dev m4 ros-noetic-eigen-conversions libccd-dev libeigen3-dev python-dev python-yaml ros-noetic-desktop-full python3-catkin-tools libpcl-dev wget git

# numerical Python
pip3 install pycddlib scipy numpy pytope matplotlib

At this point, you are ready to set up your trace-ws workspace for TRACE. If you haven't already sourced ROS, do that now:

# See ROS install instructions for your system for details if you are not used to this step:
# https://wiki.ros.org/ROS/Tutorials/InstallingandConfiguringROSEnvironment
source /opt/ros/noetic/setup.bash
cd ~
mkdir trace-ws
cd trace-ws
catkin init
git clone https://github.com/albee/trace  # might need to use SSH instead
mv trace src/trace

External dependencies (GTSAM, CasADi, Teaser++) are satisfied using the following sources:

cd src
git submodule update --init --recursive

# gtsam
pushd external/gtsam && mkdir build && cd build && cmake .. -DGTSAM_BUILD_WITH_MARCH_NATIVE=OFF -DGTSAM_USE_SYSTEM_EIGEN=ON &&  make -j2
sudo make install
popd

# casadi
pushd external/casadi && mkdir build && cd build && cmake .. && make -j2
sudo make install
popd

# teaserpp
pushd external/TEASER-plusplus && mkdir build && cd build && cmake .. -DBUILD_TEASER_FPFH=ON &&  make -j2
sudo make install
popd

Astrobee dependencies

Finally, this implementation of TRACE offers tight integration with the Astrobee simulation for visualization and systems testing. Minimal Astrobee-derived messages and base classes are therefore required; however, TRACE's core algorithms can be separated, if desired:

Follow the instructions from TUTORIAL.md for Astrobee simulation setup.

Build

This repo contains ROS-compatible packages, meant for the src directory of a traditional ROS catkin workspace.

To build, go to your trace-ws workspace :

cd ~/trace-ws
catkin build -j2

Usage

See TUTORIAL.md for a walkthrough on running TRACE in the Astrobee simulation environment!

Methods

Chaser

  • Target state estimation of Target by Chaser (angular velocity and orientation)
  • Target parameter estimation of Target by Chaser (in progress)
  • Online motion estimation of Target by Chaser
  • NLP-based motion planning method which also accounts for appendage avoidance
  • Robust tube-MPC controller for robust tracking of trajectory (plus a backup PD controller)

Target

  • Target tracking of desired tumble motion