Integrated Astronaut Control System for EVA Penn State Mars Society RASC-AL 2003
Problem Statement Future of space exploration: manned missions to Mars Exploration issues Long time delay from Earth EVAs far from home base These issues never previously encountered fully
Exploration Applications Soil and rock samples Surveying the Martian terrain Scientific observation
Spacesuit Bulkiness makes mobility difficult Lack of flexibility Gloves Hand fatigue Difficult to grasp objects Solution: Rover accompanies astronaut
Rover Assisted Exploration Rovers: tried and true Martian explorers Useful toolkit for astronauts on EVAs On-site rover control by astronauts Variety of rover control systems Joystick Trackball VR glove
Rover Control Past: Control from Earth Supercomputers Delay due to transmission over large distance Joystick control Future: On-site control by astronaut Joystick and trackball not practical VR Glove
Design Requirements Fine-tuned control No overlap between commands Efficient response to commands Simplicity and ease of training Transmission efficiency (range and power) Multitasking
Virtual Reality Gloves Simulates the environment for practical purposes Flight training Education Capabilities Six degrees of freedom Many more states than conventional controllers Feedback Data
Integration into the Spacesuit Characteristics: Mobility & Flexibility Robust Function Simple & Reliable VR Glove is small Lightweight Thin fibers Best Place to Install: Max. sensitivity to hand motions Between first and second layers
Our Solution 5DT Data Glove ActivMedia Pioneer 2-AT rover SmileCam camera Steering and camera control by VR glove
Project Timeline
Gesture Control System Data Input and Filtering Gesture Recognition State Selection Device-Specific Output
Data Input and Filtering Independent Input and Filter per hand Raw glove data calibrated to user's range of motion Exponential filter to smooth noisy data Muscle Twinges Cardiovascular pulses
Gesture Recognition Hand sensor readings 7.2e16 possible combinations! Effect of finger dependencies with imprecise control: Not this many are realistic Continuous Control: Mealy Model Discrete Control: Moore Model Hybrid Control
State Selection Each hand operates independently Certain states locked out to other hand Root state allows external operation
Device-Specific Output Translates gesture state into reasonable device output Models exist for pan/tilt cameras, motion bases, and external microcontrollers
Player/Stage Player: Robot device server Abstracts device specifics from control class Designed for networked operation from any language that supports TCP/IP Stage: Simulator for Player controllers Provides simulated environment for controller development Utilizes same binary interface as Player
Rover Navigation Uses Player's PositionDevice class Translates glove finger position and roll into rotational and translational velocities
Target Selection Translates glove gestures to control PanTilt device class Manages selection of interesting targets
A Brief Demonstration
Testing Obstacle Course requires: 1. Figure Eight 2. Arcing Turn 3. Reverse 4. Slalom Three Input Devices: Glove Joystick Trackball
Course Results User B has more training than User A Joystick is the fastest method Trackball is significantly slower
Results Analysis Results analyzed in the context of remote operations Joystick is faster, but the glove has other advantages
Future Developments Touch Sensors Force Feedback More useful user feedback Menuing Sounds Force Feedback Autonomy in Tracking and Navigation
Questions