6533b86ffe1ef96bd12ce677
RESEARCH PRODUCT
D Sensor-Based Obstacle Detection Comparing Octrees and Point clouds Using CUDA
Knut Berg KaldestadDavid A. AnisiGeir Hovlandsubject
business.industryComputer sciencePoint cloudComputer Science ApplicationsComputational sciencelaw.inventionIndustrial robotTask (computing)CUDAControl and Systems EngineeringlawModeling and SimulationObstacleRobotComputer visionCollision detectionArtificial intelligencebusinessRepresentation (mathematics)Softwaredescription
This paper presents adaptable methods for achieving fast collision detection using the GPU and Nvidia CUDA together with Octrees. Earlier related work have focused on serial methods, while this paper presents a parallel solution which shows that there is a great increase in time if the number of operations is large. Two dierent models of the environment and the industrial robot are presented, the rst is Octrees at dierent resolutions, the second is a point cloud representation. The relative merits of the two dierent world model representations are shown. In particular, the experimental results show the potential of adapting the resolution of the robot and environment models to the task at hand.
| year | journal | country | edition | language |
|---|---|---|---|---|
| 2012-01-01 | Modeling, Identification and Control: A Norwegian Research Bulletin |