Novel landmark based navigation for autonomous mobile robot

In this study, we aim to achieve an autonomous locomotion of the mobile robot in the unknown environment regardless of indoor and outdoor. An information obtained from indoor and outdoor environment are completely different. In indoor case, dense information like a wall could be obtained. However, only sparse information is obtained in outdoor case. Therefore, in this study, dense information of indoor environment is convert to sparse information and same navigation algorithm is performed in the indoor and outdoor environment. For the first step of the study, the autonomous locomotion of the mobile robot in unknown indoor environment is realized. In particular, a novel landmark construction and detection method are proposed. The landmark is generated by combined image and shape features. By combining these features, some fault of each features are filled up and robust landmark detection in various environment could be achieved. In the landmark detection step, we represent the new matching method with automatic weighting for each features. The confidences are made from image and shape indicators of degree of the characteristic. The effectiveness of the proposed landmark is verified by experiment. Moreover, we introduce the novel landmark based graph SLAM. In our method, the landmark detection is performed on each node of pose graph. Then, if the robot find the landmark which is detected once, local loop closure is generated and optimization is performed. The main advantage of the method is that we can perform graph optimization before find global loop-closure. Hence, our method can minimize global error of the graph even if the robot don't come back the place visited at once. The effectiveness of the proposed graph SLAM is verified by numerical simulation and experiment.