Monthly Technical Report

July 1999


In the technical reports for May and June 1999, we detailed the various tasks we are engaged in and the goals we have set in preparation for the next IPR meeting. In July 1999, we continued to make significant progress on these tasks and towards these goals. We concentrate mainly on the new aspects of our research in this report.

As we explained in earlier reports, our map-building strategy combines techniques we are developing for constructing polygonal models of the environment with SRI's algorithms for constructing models that are "point clouds." We are spending a significant portion of our effort on developing efficient and robust algorithms to convert from point-based models to polygonal models. In July, we implemented and experimented with fast algorithms for matching and merging different partial polygonal models of the same environment. We have obtained promising preliminary results.

We are also continuing our progress on target-tracking. Our main focus now is to implement a "smart" controller for the robot that will take high-level motion commands from the target-tracking planner and convert these commands into smooth robot and camera motions. We are experimenting with various strategies for generating such smooth motions. To compute such motions, we combine several aspects into account such as the relative positions of the robot and target, the path the planner desires to move the robot on, and the trajectory and velocity of the target. We generate the motion of the robot and camera using fast and robust numerical techniques.

We have been facing some hardware problems on our robots of late. The hard drives on two of our robots have crashed a few times. We conjecture that these crashes are caused by excessive heat inside the robot casing. We have solved this problem on one of our robots by drilling a series of holes around the robot body and installing fans near these holes to create some circulation. Since we made this modification, the temperature inside this particular robot has remained low and its hard drive has not crashed. We are planning to similarly modify the other robots.