Enabling Easier Programming of Machine Learning Algorithms on Robots with oneAPI Toolkits.

Abstract

This work shows that it is feasible to solve large-scale decision-making problems for robot navigation in real-time onboard low-power heterogeneous CPU+iGPU platforms. We can achieve both performance and productivity by carefully selecting the scheduling strategy and programming model. In particular, we remark that the oneAPI programming model creates new opportunities to improve productivity, performance, and efficiency in low-power systems. Our experimental results show that the implementations based on the oneAPI programming model are up to 5× easier to program than those based on OpenCL while incurring only 3 to 8% overhead for low-power systems.