Statistical characterization of sensors for sensor fusion

[摘要] It is standard practice in advanced telerobotic systems to use a detailed model of the operating environment to plan tasks and drive simulations. How well these systems operate is highly dependent upon the accuracy of this model. As objects are moved by manipulators, inaccuracies in the world model will accumulate, progressively reducing the overall system utility. To compensate for errors in the world model, sensory information on the location of objects within the operating area must be used to drive world model updates. Every sensory process (e.g. laser range-finding) provides information of limited accuracy, and this accuracy function must be taken into account during the data fusion. We measured and characterized sensor inaccuracies for video cameras, laser range-finders and kinematic positioning information, and used them as the foundation of a sensor fusion system to produce more accurate position measurements of objects in the workspace. By studying the resulting fused sensor error characteristics, we show how sensor sub-systems for telerobotic systems can be designed, and how fused sensor information can be used to increase manipulator accuracy.

[发布机构] Rice University

[效力级别] Electrical engineering