Constrained 6DOF localisation for autonomous vehicles

Autonomous navigation for vehicles in non-trivial environments requires fast and precise localisation. Vehicle control based on erroneous localisation can lead to motion from which it is costly to recover. Autonomous vehicles typically operate in 3D environments and require 6DOF localisation, which is computationally costly. This paper reports the development of an accurate 6DOF particle filter based localisation system which is suitable for localisation of autonomous vehicles operating in 3D environments. Fast, accurate localisation is achieved by constraining the 6DOF search space by physical vehicle limitations, and by the use of Graphics Processing Units (GPU´s) to enable the use of dense 3D internal representations of the environments in the localisation process. Experimental results of the localisation system controlling autonomous navigation are presented, with an average localisation error of approximately 10cm.