Building Consistent Laser Scan Maps

The paper presents an approach to generate consistent maps in real-time using a laser range sensor. Its application scenario is the Bremen Autonomous Wheelchair “Rolland” that is developed as an autonomous transport vehicle for hospitals. The paper consists of two parts: in the first one, a laser scan matching method is presented that in several aspects improves an algorithm originally proposed by Weiß et al. [14]. In the second part, a mapping approach is introduced that is based on the scan matching method. Using a laser scanner mounted on the wheelchair, it generates consistent maps, i.e. maps that do not have discontinuities resulting from the accumulation of errors in the mapping process. As such accumulations cannot be avoided, the resulting errors are corrected whenever the wheelchair returns to areas already mapped. An important property of the presented approach is its real-time capability. The mapping is performed on the fly and thus, the resulting maps can immediately be employed for self-localization.