INL’s “guarded motion” (an example of reflexive teleoperation) capabilities exploit several different sensors (including scanning laser, infrared, sonar, tactile, inertial, and tilt), fusing available perceptual data into regions that represent the ability of the robot to move safely in a given direction. The algorithm continuously calculates an “event horizon” representing the last possible moment for the collision avoidance behavior to successfully intervene at the current speed. By calculating this event horizon many times each second, the robot can smoothly scale down its speed as a function of congestion without losing freedom of motion. In terms of portability between robots, the event horizon provides an implicit means to adapt the guarded motion to different deceleration models. When a full stop is required, use of the event horizon ensures that the robot comes to a halt at the same distance from an obstacle regardless of its speed. INL’s remote-operation studies with experienced and novice participants have shown that this predictability improves the operator’s trust of and ability to use the system.
The guarded motion capabilities are also based on a continuous assessment of battery power and the reliability of sensor data. The robot uses this assessment to provide the operator with a measurement of its health that based on remaining power, sensor reliability, and other critical information. Built on top of the guarded motion, the robot’s obstacle avoidance capabilities allow it to circumvent obstacles and extract itself from box canyons. It may refuse to undertake a task it cannot safely accomplish. For instance, the robot may not to exceed certain speeds if its laser data become invalid and it is forced to rely only on sonar. The end result is that an operator may issue motion commands without fear of harming the robot, which greatly accelerates the speed and confidence with which the operator can accomplish remote tasks.
- Contact:
- David Bruemmer, Send E-mail