FöRFATTARE
Larsson, Ulf

DATUM
1999-11-08

INSTITUTION
Systemteknik / Robotik och automation

SAMMANFATTNING
The problem studied in this thesis is control of robots using information from optronic sensors. With a laser range sensor onboard the robot the relative distance to an object can be measured. This allows the robot to operate in a partial unknown environment. One complication with non-contact sensors is the assiciation problem, i.e. finding the correspondence between the measured signals and each object in an unstructured workspace. This problem is addressed in the algorithms presented for using measurements. The approach is model based with respect to sensor errors, uncertainty in association and robot dynamics. Results includes:

* The range weighted Hough transform (RWHT) was developed as a robust algorithm for extracting plane surfaces in scans made by a range measuring laser. The peaks in the RWHT are used during navigation of mobile robots and for automatic generation of maps.

* A modelling study is made toward improving mobile robot performance using rate gyros. The main limitation seems to be an unknown lateral velocity. The special kinematic for navigating articulated vehicles is also studied.

* An hydraulic arm, without joint angle encoders, is controlled with feedback from measured directions to retro-reflectors. An eye-in-hand range camera is studied with respect to gripping and for reducing vibration.

Considering the fast development in computing and laser based sensing there is a large potential for novel application like robots in contaminated areas, in mining, in forests etc. The model-based approach in well suited for industrial requirements on predictable properties and self-monitoring of control and information systems.

ISSN 1402-1544 / ISRN LTU-DT--99/29--SE / NR 1999:29