Arc-Line Features based Two-Dimensional SLAM with Partial Compatibility Algorithm and Summing Parameters

Abstract

To realize simultaneous localization and mapping (SLAM) with different types of features, a mutually converted arc-line segment-based SLAM algorithm by distinguishing what we call the summing parameters from other types, and a novel data association algorithm named as partial compatibility algorithm are proposed. These redefined parameters are a combination of the coordinate values of the measuring points. Unlike most traditional features-based SLAM algorithms that only update the same type of features with a covariance matrix, our algorithm can match and update different types of features, such as the arc and line. For each separated data set from every new scan, the necessary information of the measured points is stored by the small constant number of the summing parameters. The geometric parameters of arc and line segments are extracted according to the different limit values but based on the same parameters, from which their covariance matrix can also be computed. If a stored map feature matches with a new extracted segment successfully, two segments can be merged as one whether the features are the same type or not. The mergence is achieved by only summing the corresponding summing parameters of the two segments.

In data association part, a partial compatibility algorithm obtaining a robust matching result with low computational complexity is proposed. This method divides all the extracted features (arc and line segments) at every step into several groups. In each group, the local best matching vector between the extracted features and the stored ones is found by joint compatibility, while the nearest feature for every new extracted corner is checked by individual compatibility. All these groups with the local best matching vector and the unit matching pairs of each new extracted feature are combined, and their partial compatibility is checked by using the branch and bound method with linear computation time. To validate the robust matching result and low computational complexity of the partial compatibility algorithm, the experimental results in an indoor environment with extracted corners are presented by comparing with individual compatibility nearest neighbor and joint compatibility branch and bound. In addition, three SLAM experiments with partial compatibility algorithm and summing parameters in indoor environments including different types of objects are done to demonstrate the robustness, accuracy, and effectiveness of the proposed arc-line based SLAM method. The data set of the MIT CSAIL Building is used to validate that our methods have good adaptability.