GPGPU Cost Function Generation with Subregion Optimisation

A slowed down simulation of a GPGPU based cost-to-go function that utilises subregion optimisation. The occupancy grid is broken into subregions and only active subregions are requested for evaluation on GPU. Active subregions are shown in white. This method is used to find a balance between the burden of per pixel micromanagement against simply requesting the whole occupancy grid being requested for evaulation each iteration resulting in a large number of wasted calculations.

Category:

Science & Technology

Tags:

• gpgpu
• dynamic programming
• cost-togo function
• path planning
• robotics