Coverage for Mobile Robots with Uncertainty

coverage;robotics;mobile robot;calibration;virtual robot

Das, Colin E. ; Bretl ; Timothy W.

关键词: coverage;    robotics;    mobile robot;    calibration;    virtual robot;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/24229/Das_Colin.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The classical problem of robot coverage is to plan a path that brings a point on a robot within a fixed distance of every point in the free space. This problem is solved under the assumption that a robot follows the path exactly. However, a robot with uncertainty in sensing and actuation is not guaranteed to follow the path exactly and thus is not guaranteed to cover the free space. A coverage planner, rather than producing a path, should produce a feedback policy, but it is unclear exactly what performance it is to achieve. To make the problem concrete, we formulate a “probably approximately correct” measure of performance that captures the probability 1 - e of covering a fraction 1 - d of the free space.We apply this measure to a simple system in which a robot is commanded to follow a square wave path in a rectangular region. We show that following a square wave path with passes parallel to the short axis performs better than following a square wave path with passes parallel to the long axis. This inspires a new strategy in which a square wave path is steered throughout the free space. We design a feedback policy where the square wave path is treated as a larger, virtual robot with coverage implement size equal to the size of the square wave. The virtual robot tracks a global path and generates a local path that the robot tracks. Coverage planning is then simplified by decoupling local and global coveragestrategies. We describe in detail the coverage planning algorithms for the robot and the virtual robot, specifically the calibration, estimation, and control processes.For the particular case of the virtual robot operating along the boundary, we introduce a method of simultaneous coverage and calibration. For robots with boundary sensors that trigger when entering or leaving the free space, like the ones we consider, we show that it is possible to estimate the robot odometric parameters based on the difference in time between boundary crossings. Through an analysis of observability, we design a sequence of switchback trajectories that produce robust parameter estimations. This algorithm is implemented both in simulation and hardware experiment.