The user interface supports several driving modes. In the direct teleoperation mode, the human has full control over the rover - almost all safeguarding is turned off. Direct teleoperation is necessary when the rover gets into situations where the software would otherwise prevent motion. For instance, there may be occasions where the pitch limits must temporarily be exceeded to drive the rover out of a crater. This mode is reserved for experienced drivers in exceptional situations.
On the other side of the spectrum, in the autonomous mode the software system has complete control over the robot, choosing the direction to travel based on the stereo input. While we have done some experiments in which we add input from a goal-directed planner, to bias the robot in particular directions, the majority of our experiments have consisted in letting the robot "wander" autonomously, while avoiding obstacles.
The third mode, safeguarded teleoperation, is seen as the standard way in which the lunar rover will be operated. In this mode, input from the human and the obstacle avoidance planner are combined: the user presents a desired direction to travel, and the obstacle avoidance planner can veto it, causing the robot to refuse to travel in that direction. The idea is that the software safeguards should prevent the user from damaging the rover, but not otherwise interfere with the control. Basically, this mode is implemented by having the arbiter combine input from the direct teleoperation mode and the autonomous mode. In addition, if the human chooses not to provide input, the rover will navigate autonomously. In this way, operator fatigue can be reduced by letting the robot operate on its own when it is in benign terrain, while still enabling the user to take over control at any moment.
LRD Navigation Group - firstname.lastname@example.org (last updated in March 1995)