Robotic Space Simulator: Design and Characterization of a Test and Evaluation Platform for In-Space Robotics

In this paper we introduce the Robotic Space Simulator (RSS), a new testbed for in-space robotics using two 7-DOF Gough-Stewart platforms. The RSS was developed to simulate a spacecraft with a robotic manipulator approaching, grappling, manipulating, and decoupling from another spacecraft, fusing accurately simulated centroidal dynamics with real force-sensorized contact interaction and realistic visuals. RSS addresses many of the deficiencies of other physical simulation tools by operating in SE(3) with high payload capacity, increased precision and reachability, and without additional requirements such as waterproofing of test articles. The RSS was designed to enable the testing of full-scale flight articles with a workspace that permits testing of multiple spacecraft and full-contact interactions between them via robotic manipulator. In this paper, we outline critical design criteria to meet these requirements such as payload capacity and reachability, and sensor range and resolution. The RSS uses force-torque sensors on each platform to simulate full-contact microgravity dynamics between the spacecraft. The sensor design prioritizes measurement range, so that the effective platform payload is not limited by the sensor. With that constraint, we next prioritize sensor selection based on their measurement resolution to maximize sensing capability and simulation fidelity. In addition to design optimization, we present initial evaluation of the disturbance characteristics of the platform as measurable by the incorporated sensors. Additionally, we evaluate the inverse and forward kinematics solving capabilities and their impact on the Jacobian-based feed-forward controller. Next, we present the calculated workspace of each platform including their additional axes. Overall we find the physical design, layout, and sensing capabilities of the RSS to be suitable as a test and evaluation simulation platform for space robotics. We find there is no significant disturbance due to the platform actuators which minimizes the need for signal filtering. We find that the Jacobian is well-conditioned throughout the reachable workspace of the platform, and the forward kinematics numerical solver is 100% successful in an average of $19.2 \mu\mathrm{s}$, which is well under the 4 m s period of the control loop. Finally, we present the workspace of the RSS platforms including their additional axes. We calculate the effective workspace of the simulator to be approximately $250\ \mathrm{m}^{2}$.