Scheduling and path planning for computational ferrying

Abstract

Mobile devices are expected to perform complex tasks exceeding their current hardware capabilities. Traditional cloud-based solutions are often useful, but fail in adversarial environments with limited communication connectivity. This can be exemplified in battlefield scenarios where soldiers may require access to large processing capabilities, without sacrificing their own mobility. In this paper, we extend previous work on computational ferrying, where Mobile High Performance Computers (MHPCs) physically move the necessary hardware into the proximity of mobile devices. Our extension proposes several improvements: 1) Path planning, which is used for reliable a priori estimation of distances between locations, yielding more accurate scheduling; 2) Collision checking, which permits robots to carry out the role of MHPCs; and 3) Prioritization, allowing operators to assign weights to tasks. In this paper, our algorithms are implemented and tested, with a comparison to previous work.