Skip to main content
FARO LAB
Field and Aerial Robotics Laboratory
WVU Home

Parallel and cloud computing for long-term robotics

Amazon Project


This project will investigate how parallel and cloud computing can be used in the development of long-term autonomous mobile robots, also known as long-running robots. These robots must operate for long periods in unstructured and dynamic environments. Examples of long-running robots are self-driving cars, delivery vehicles, space exploration rovers, and agricultural robots.

The project will concentrate on mapping, a never-ending task in long-term robotics, and motion planning, an important need for task completion and safety. While mapping generates large amounts of data, motion planning must be executed as fast as possible.

Our main goal is to develop an open-source computational architecture for long-running robotics that seamlessly integrate on-board parallel planners running on CUDA-enabled computers with mapping strategies executed in a cloud environment.

Architecture for the proposed approach: Locally a robot creates a local map using its sensors. This map is used for motion planning and is sent to the cloud where it is stored and eventually combined with other maps. The motion planner uses CUDA.

This architecture will be tested using Amazon EC2, AWS RoboMaker, NVIDIA Jetson Nano and two different kinds of long-running robots: a tether-powered drone operating in a warehouse and a self-recharging ground robot working in an office-like environment.