Improving the obstacle detection and identification algorithms of a laserscanner-based collision avoidance system

Advanced driver assistance systems represent considerable progress for improving the safety of vehicles on the road. One group of these systems is based on the detection of obstacles. If positive results are to be achieved these systems must be capable of taking a coherent decision as to which situations involve hazard, in order to avoid false alarms that lead to actions that are not only wrong but disturbing to other road users and cause the driver to lose faith in the system. This paper presents some algorithms to improve those already existing for detecting, identifying and characterising obstacles by means of a laserscanner. The major innovations are: (1) fusing the information from the laserscanner with a positioning system while taking account of the quality of the latter; (2) the criteria for locating obstacles (segmentation process), overcoming the limitations of other approaches that ignore the influence of the obstacle’s orientation; (3) the method of defining the characteristic axes of the obstacles, without resorting to tolerance values that are difficult to adjust or reducing the influence of distance measurement errors of the laserscanner. The algorithms were tested with on-track tests using a Sick LRS 1000 long-range laserscanner with satisfactory results being attained that were an improvement on those provided by other methods.

Research highlights
► Novel algorithms for detecting obstacles by means of a laserscanner are proposed.
► A positioning system is used to focus the laserscanner information.
► Novel criteria for locating obstacles improve other approaches.
► The algorithm defines the characteristic axes of the obstacles.