가정용 청소 로봇을 위한 청소 주행 알고리즘에 관한 연구

Abstract

Although cleaning robots have been increasingly popular in home environments, their coverage rate and performance have not been very impressive to their users so far, thus often hampering their user acceptance. Complete coverage usually mandates the robot to have a sophisticated navigation system for precise position estimation or localization as well as a precise map. To this end, the robot must be equipped with mid-to-long range sensors such as high-cost 2D laser range finders and/or vision sensors whose data processing tends to be computation intensive ？ thus not suitable for a home environment. Furthermore, the well-known SLAM algorithms, due to their complexity, cannot directly be used for low-power embedded systems. This paper presents a novel integrated coverage strategy for low-cost cleaning robots, yet demonstrating a respectable coverage performance in any unknown environment. We present an efficient algorithm that can cope with hardware limitations, ranging from low computational power to numerous sensing problems arising from limited range and poor directionality, detection uncertainty, and large measurement error. To facilitate a viable solution that can cope with these limitations, we first make two assumptions on the home environment ？ rectilinear and closed, which are met in most of our home environments. The first is justified by noting that the major structures of the indoor environment can be represented by sets of lines which are parallel or perpendicular to each other. The second stipulates that the robot can fully cover the target area by following the walls. In particular, this orthogonality assumption turns out to be very crucial in assisting us to localize and plan a coverage path in spite of minimal sensing of the environment. Next, in order to effectively circumvent poor localization (low precision positioning), we decompose the space into sectors to maximize coverage, with each sector being small enough to have reasonable localization accuracy within itself. This is very important to achieve a good coverage performance. Overall, the final outcome is a novel online coverage strategy that simultaneous performs exploration, incremental sector creation, and sector cleaning with robot localization, with the intention of maximizing performance with minimal sensing. Both simulation and real world experiments validate the efficacy of our approach.