An Approach to Low-Cost Mobile Indoor Robotics using an Edge Network Computing Paradigm

Abstract

In this project we develop a novel approach for computational offloading in robotics. Traditionally, computational offloading has been synonymous with cloud computing, or running all computation on a remote server. We show that this is both relatively expensive and slow, and we propose an alternative offloading paradigm by which computation is moved to local, userowned computers. This approach can reduce computational costs, network bandwidth usage, and data transmission latency. In this report, we examine and quantify the potential benefits of such a system, identify the challenges involved in extending the existing cloud computing framework to local computers, and propose a new system — the Load Manager — which handles the additional complexity incurred by interactions with these local computers. Finally, we detail our implementation of a Load Manager in the context of a small ground-based robot running autonomous navigation software, and analyze the system’s overall performance.