Technologies for low-bandwidth high-latency unmanned ground vehicle control

Automation technology has evolved at a rapid pace in recent years; however, many real-world problems require contextual understanding, problem solving, and other forms of higher-order thinking that extends beyond the capabilities of robots for the foreseeable future. This limits the complexity of automation which can be supplied to modern unmanned ground robots (UGV) and necessitates human-in-the-loop monitoring and control for some portions of missions. In order for the human operator to make decisions and provide tasking during key portions of the mission, existing solutions first derive significant information from a potentially dense reconstruction of the scene utilizing LIDAR, video, and other onboard sensors. A dense reconstruction contains too much data for real-time transmission over a modern wireless data link, so the robot electronics must first condense the scene representation prior to transmission. The control station receives this condensed scene representations and provides visual information to the human operator; the human operator then provides tele-operation commands in real-time to the robot. This paper discusses approaches to dense scene reduction of the data required to transmit to a human-in-the loop as well as the challenges associated with them. In addition, the complex and unstructured nature of real-world environments increases the need for tele-operation. Furthermore, many environments reduce the bandwidth and increase the latency of the link. Ultimately, worsening conditions will cause the tele-operation control process to break down, rendering the robot ineffective. In a worst-case scenario, extreme conditions causing a complete loss-of-communications could result in mission failure and loss of the vehicle.