XPS: FULL: DSD: Parallel Motion Planning for Cloud-connected Robots

University Of North Carolina At Chapel Hill, Chapel Hill NC

Investigators

Abstract

Robots are entering new domains, from self-driving vehicles on real-world streets, to autonomous aerial vehicles for package delivery, to assistive robots helping people with disabilities in their homes with daily activities. Autonomous robots in these domains often need extensive computational resources for motion planning, which involves computing a safe motion for a robot through an environment that avoids obstacles and accomplishes the task. To enable low-power mobile robots to achieve their full potential, new algorithms and software frameworks are needed that fully leverage parallel computation and the warehouse-scale computing available via the cloud. This project aims to develop a new software framework for motion planning for cloud-connected robots that effectively parallelizes motion planning and distributes the computation across the robot's embedded multicore processor and multiple cloud-based compute servers. This research combines ideas from multiple areas of computer science and engineering, including robotics, parallel algorithms, high-performance computing, motion planning, and cloud computing. Locally on the robot, the project parallelizes traditional motion planners to fully leverage low-power, embedded multicore processors. Simultaneously, the project enables the robot to request computation time from cloud-based resources to significantly increase the computing power available for the motion planning task, and thus increase the responsiveness and quality of motion plans. The new algorithms and software aim to enable robots to autonomously complete tasks in new domains where the challenge of motion planning is currently prohibitive, broadening the applicability of robots to new societally-relevant domains. The concepts and software developed in this project are being integrated into undergraduate and graduate courses taught across topics ranging from robotics to high-performance computing. Another goal of the project is to create fun, hands-on, interactive demonstrations using cloud-connected robots to inspire children to consider STEM fields.