Cloud-Based Services for Cooperative Robot Learning of 3D Object Detection and Recognition

摘要

Some of the problems preventing the widespread adoption of advanced service robots are (1) cost, (2) power consumption, (3) perceptual capabilities, (4) knowledge management, (5) reasoning capabilities, and (6) compute power. Improvement along these dimensions will make adoption of service robots with advanced capabilities much more widespread. We propose the use of real time cloud computation as one means to enable these improvements. This paper presents a case study on the use of a cloud computing platform to support robotics applications, allowing robots to offload heavy compute tasks such as machine vision to cloud infrastructure. We specifically aim to give a widely distributed group of robots the ability to learn 3D objects cooperatively for detection and recognition. Participating robots can share their knowledge with others via our cloud-based services. Experiments with a proof of concept prototype demonstrate the feasibility of the use of cloud platforms to deliver improved perceptual, knowledge management, and reasoning capabilities while keeping the cost and power consumption of the robot low.