Motion planning is determining how a robot should move to achieve a goal; for example, a to move the arm of a robot to a desired destination avoiding collisions with any obstacles.
Decisions must be made in real time. The robot sensors have to analyze its surrounding environment, and calculate an optimum path from source to destination while avoiding obstacles.
Until now most solutions were based on GPUs. Typical decision making time for GPU based solutions was on the order of hundreds of milliseconds. The FPGA based solution explored on this paper can reduce this time to the order of tenths of milliseconds, namely, around a thousand times smaller.
One typical solution for motion planning is building what is called a PRM, i.e., a Probabilistic RoadMap. The PRM contains all possible paths from source to target. According one study, the collision analysis between the robot and the obstacles amounts to 99% of the needed compute power for the motion planner sampled on said study.
In this new approach, the collision detectors are implemented using blocks inside an FPGA. The collision detectors work in parallel, and it is this parallelism who gives the thousand-fold advantage in planning time.
For more details, please refer to Duke University’s paper on this issue.