Edge Computing-Based Collaborative Vehicles 3D Mapping in Real Time

Cooperative vehicles are better able to detect the environment and self-localize than a single vehicle. Cooperative vehicles can quickly cover the entire environment by communicating and cooperating with each other and can also reduce localization and mapping error by merging the cooperative vehicle information from observation and navigation. In this paper, the authors propose a novel algorithm for an effective solution of navigation and mapping for cooperative vehicles in an unknown environment. The authors present an improved centralized and collaborative monocular simultaneous localization and mapping (CCM-SLAM) approach. The proposed algorithm can accurately compute the transformation matrix for cooperative vehicle maps and reduce the communication delay, data loss among vehicles and decrease the bandwidth demand. The quaternion and credibility similarity transformation (QC-Sim(3)) method the authors proposed is used to accurately merge the matched maps and accomplish loop closures. The sending messages at variable frequencies (SMVF) method the authors proposed and an improved detection and resending lost data (I-DRLD) method the authors proposed can improve the accuracy of pose estimation. SMVF solves the time-delay problem by sending messages to the vehicles at flexible frequencies while I-DRLD detects and resends the lost data. The authors also adopt Intra-frame Feature Compression (IFC) to decrease the bandwidth demand in the process of the transmitting data. The experiments demonstrate the superiority of our proposed algorithm compared with the state-of-the-art methods.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01760588
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Nov 20 2020 12:24PM