Indoor Autonomous Localization Method Based On Multi-sensor By Using Data Fusion

With the development and further research in AI(Artificial Intelligence) field, SLAM(Simultaneous Localization and Mapping) has a vast application prospect in aspects of human-computer interaction, exploration of unknown environment(like space, ocean and underground), disaster relief and reconstruction as very hot-heated issues in the AI field. There are two key technologies in SLAM, one of them is the autonomous localization which is localizing oneself in an unknown environment, and the other one is recreating a map of this environment. These two tasks are coupled to each other: an accurate localization is benefited to mapping, and a good map is crucial to localize oneself. From previous works, we found that the majority hardware system is monocular vision sensor like camera. To solve the problem that monocular camera can’t provide depth information directly (need a complex calculation by EKF), the author proposes a novel feature-level data fusion based indoor autonomous localization method. The proposed method is applied in inactive multiple-reference unknown indoor environment. The main contribution of this paper is as follows:Firstly, after reviewed the previous works around SLAM, the author reproduces the autonomous localization method based on monocular sensor. The results of the work show that: this method can’t provide an accurate localization as a result of the incomplete information (without depth information). So, our work focus on multi-sensor of different ’types’based indoor autonomous localization.Secondly, an intelligent platform named’smart car’is constructed by51MCU, drive circuit, router with OpenWRT for indoor autonomous localization. User can control the motion parameters as moving direction, speed, acceleration and so on. They also obtain the information from the sensors (like camera) on the car. With the implementation of the algorithm of feature detection, EKF prediction, data association and pose estimation, a ’smart car’ based on monocular indoor autonomous localization method is implemented.Finally, a feature-level data fusion method is proposed to solve the problem that monocular camera can’t provide depth information directly and fleetly. Images from camera and depth information are measured in the proposed feature-level data fusion method. Compared to the raw data, the fused information shows a satisfactory performance in the experimental results. So, a novel indoor autonomous localization basing multi-sensor feature-level data fusion is proposed. The processes of the localization method are shown as follows:simultaneous camera pose and multi-feature information of fused data can be calculated through each other. From this, a cycle of the pose estimation is complete. The experimental results prove the fact that the accuracy of the estimation can be improved through this feature-level data fusion method.