Supervising Remote Humanoids Across Intermediate Time Delay * Kimberly Hambuchen, William Bluethmann, Michael Goza, Robert Ambrose Automation, Robotics and Simulation Division NASA Johnson Space Center Houston, TX 77058 {hambuchen, bluethmann, sgoza, ambrose }@jsc.nasa.gov * NASA and the Mobile Autonomous Robot Software program in the DARPA Information Processing Technology Office (IPTO) sponsored this work Kenneth Rabe Mark Allan Planning Software Systems Group Intelligent Robotics Group Jet Propulsion Laboratory NASA Ames Research Center Pasadena, CA 91109 Moffett Field, CA 94035 kjrabe@jpl.nasa.gov mallan@email.arc.nasa.gov The President’s Vision for Space Exploration, laid out in 2004, relies heavily upon robotic exploration of the lunar surface in early phases of the program. Prior to the arrival of astronauts on the lunar surface, these robots will be required to be controlled across space and time, posing a considerable challenge for traditional telepresence techniques. Because time delays will be measured in seconds, not minutes as is the case for Mars Exploration, uploading the plan for a day seems excessive. An approach for controlling humanoids under intermediate time delay is presented. This approach uses software running within a ground control cockpit to predict an immersed robot supervisor’s motions which the remote humanoid autonomously executes. Initial results are presented. Index Terms – control over time delay, space exploration, remote robots, task assistant, sensory egosphere, autonomous tool use. I. I NTRODUCTION In January 2004, the President’s Vision For Space Exploration laid out the strategic plan for exploration of the solar system [1]. Early stages of the plan call for robotic missions to the Earth’s moon to demonstrate new technologies and to initiate work on operations prior to the arrival of human astronauts. Functions to be completed during these precursor missions will likely include mapping the lunar surface, precision landing, environmental monitoring, communications network setup, infrastructure build-up and in-situ resource utilization [2]. The machines and robots to complete these tasks will inevitably be varied in shape and form. Within this portfolio of machines, there will be humanoids that have manipulative capabilities. These dexterous robots will likely have some degrees of autonomy but may need to be supervised by humans. Due to the limited number of astronauts and their full schedules while in space or on the lunar surface, ground personnel will likely need to remotely supervise these robots. Based on the speed of light, the round trip delay between issuing a command from Earth to the moon and seeing any result from that command is on the order of 1.5 seconds. A round trip time delay closer to 10 seconds is possible with data being routed through various satellites. Even under a 1.5 second time delay, bilateral control without compensation causes instabilities in a robot; bilateral control under delays up to 10 seconds would be very challenging. Solutions for controlling remote robots over time delay usually fall into one of four methods, or a combination of these methods: 1) “move-and-wait”, 2) bilateral control stabilization, 3) predictive displays and 4) supervisory control. From a control systems perspective, the simplest method is the “bump and wait” technique, i.e. a teleoperator inputs small commands then waits for the motion to settle. The “bump and wait” solution can be effective, though the teleoperator wastes a large amount of time by sitting idle. Astronauts on the Space Shuttle and International Space Station (ISS) employ the “bump and wait” technique when berthing large payloads with remote manipulator systems, although no time delay occurs between command and feedback. A significant amount of effort has been put forth in stabilizing bilateral control of manipulators across time delay. The seminal works in this area were published by Anderson and Spong (scattering theory) and extended by Niemeyer and Slotine (wave variables) [3, 4] and Bejczy and Kim (shared compliance control) [5]. In 1999, bilateral control of sliding and peg-in-hole tasks were successfully completed across a 7 second time delay during the Engineering Test Satellite 7 (ETS-VII) experiment [6]. This controller employed a modified PD controller for the bilateral control. Predictive display methods immerse the teleoperator in an environment with solid or wire-frame virtual models of the remote location overlaid onto live video. The teleoperator can view past, present and future states of the remote robot. Past views are represented by delayed video. Present views are found in predictions of the current state of the robot based on past commands from the previous time delay period and a model of the remote environment. The future is represented by the commands currently leaving the ground. This method augmented with intelligence on-board the remote robot was
