FLOATS: Flexible Light-weight On-orbit Actuator Simulation

A simulation framework for free-flying soft robots in microgravity.

---

About the Project

FLOATS facilitates realistic simulation of locomotion, manipulation, and planning for deformable on-orbit systems. We demonstrate grasping, free-floating object reconfiguration, and mobile manipulation tasks as a baseline for research on soft robotics planning and control in microgravity.

Features

• PyBullet Simulation Environment:
  • Loads ISS environment and mission objects.
• Manipulator Control:
  • Load and configure a soft continuum manipulator.
  • Apply torque control and set contact properties.
• Path Planning Algorithms:
  • RRT
  • RRT*
  • RRT-Connect
  • A* search
• Grasping Strategies:
  • Apply hard-coded trajectories for inhand-manipulation,
  • Execute torque-based grasps with dynamic adjustments.

Getting Started

Test inhand manipulation

You can perform inhand manipulation by grasping a cargo transfer bag and rotating it by 180 or 360 degrees. The procedure consists of three phases: grasping, releasing with induced spin, and timed regrasping. Therefore, we use a trajectory defined in "trajectory_inhand_manipulation", that you can adapt to also manipulate different objects.

python test_inhand_manipulation.py

Test grasping

We provide a grasping benchmark with six space-relevant objects. These include a camera, a cargo transfer bag, a sample tube, a rover wheel, a grease gun, and a crew bag. However, only cargo transfer bag, rover wheel and crew bag can be grasped sucessfully from all six directions.

python test_grasping.py

Test path planning

We simulate module-to-module transfer of a camera from the Kibo ELM-PS module to the Cupola module. The experiment allows for four standard planning algorithms: A*, RRT, RRT*, and RRT-Connect.

python test_planning.py

Prerequisites

SOMO: https://github.com/GrauleM/somo.git

Ubuntu 16.04 and Ubuntu 18.04 with Python 3.6.9
Ubuntu 20.04 with Python 3.6.9, 3.7.9 and 3.8.2
Windows 10 with Python 3.7 through Anaconda