Jaakko Karras: Improving Inertial Sensors on the OVERBOT

Improving Inertial Sensors on the OVERBOT

Jaakko Karras worked under the guidance of Professor Gabriel Elkaim and John Burr to restore software and improve inertial sensors on the Overbot autonomous vehicle. The Overbot consists of a six-wheeled Polaris Ranger utility vehicle, modified with an assortment of sensors, actuators and computers which perform obstacle avoidance and path finding tasks. The primary sensors employed on the Overbot include a GPS unit, inertial measurement sensors, and a laser range finder. A former DARPA Grand Challenge entrant, the Overbot was donated to the UCSC Autonomous Systems Lab  during the fall of 2005.

The first phase of this project involved restoring the Overbot codebase to a working state, so that it could be expanded to investigate new path planning algorithms. This required becoming familiar with the existing software and the QNX operating system, as well as tracking past changes and bugs that had rendered the codebase non-functional. A series of shell scripts was also written to facilitate with future software installations on the Overbot.

With the software restored to a working state, Jaakko focused on reducing temporal drift in the Overbot's inertial sensors. The Overbot uses inertial sensors to correct for errors in its GPS readings. The existing sensors, however, have a tendency to drift over time, thereby producing erroneous estimates of the vehicle's position. To mitigate this problem, the existing inertial unit will be replaced with two high-precision modules, interfaced to the Overbot computer through a dsPIC33F microcontroller. Jaakko produced both the software and hardware for communicating with one of the two sensors, a fiber-optic gyroscope. The resulting package reduces drift in heading measurements from approximately 200 down to 18 degrees per hour.