An Assessment of Vision Systems for Tele-Operated Ground Robot Navigation

摘要

Many environments are hazardous to humans, requiring the use of robotics for certain tasks. These tasks, including navigation, are often conducted using a vision system for robot tele-operation. This paper aims to assess the effectiveness of various vision systems, including: in-situ monovision, aerial monovision, combined monovision, and stereovision, in providing the operator with sufficient information for tele-operated ground robot navigation. A simulated tele-operation experiment was designed and users were tasked with the objective of locating and navigating to a given target in a debris field. User performance was recorded based on the following metrics: total navigation time, total move commands, invalid move commands, and time required to make a navigation decision. The vision systems were assessed based on the system’s ability to assist the user in performing the task. An analysis of the results was conducted and it is proposed that only a combined monovision system or a stereovision system is sufficient for tele-operated ground robot navigation tasks. Nomenclature b = baseline f = focal length Pl = pixel, left image Pr = pixel, right image α = significance level p = p-value, probability I.