Taiyuan Ligong Daxue xuebao (Sep 2023)
Efficient Data Compression for CVSLAM Based on Arithmetic Coding in ORB-SLAM2 Framework
Abstract
Purposes Collaborative Visual Simultaneous Localization and Mapping (CVSLAM) has attracted more and more researchers’ attention in the field of robotics owing to its low cost of required sensors, the ability to acquire rich environmental information, and its rapidity and flexibility. Realizing the efficient transmission of image information is one of the key problems that need to be solved to improve the efficiency of CVSLAM map building. In multi-machine cooperative operation, data transmission between robots is the most important part of their work, and data sharing is often affected by the communication bandwidth, so the study of efficient data processing methods has become a crucial part. Methods On the basis of the centralized collaborative ORB-SLAM2 framework, the problem of data transmission from individual robots to the central station is studied, and a general method of efficient data transmission based on feature compression is explored. In particular, a method based on arithmetic coding for differential coding of ORB features and compressed transmission is investigated. The method evaluates the amount of data to be transmitted after compression according to different coding modes of the features and selects the mode that minimizes the amount of data transmission to encode the features. Findings Without affecting the overall map building effect, the adopted method significantly reduces the amount of data during data transmission in the CVSLAM system, and at the same time shortens the data transmission time. Conclusions The experimental results based on the KITTI dataset show that the communication method with compression coding can effectively reduce the amount of data transmission and increase the transmission speed, which is an effective means to improve the efficiency of CVSLAM communication.
Keywords