# Improving Autonomous Vehicle Mapping and Navigation in Work Zones Using Crowdsourcing Vehicle Trajectories

**Authors:** Hanlin Chen, Renyuan Luo, Yiheng Feng  
**arXiv:** [2301.09194](https://arxiv.org/abs/2301.09194)  
**Date:** 2023-01-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2301.09194v1/

## Key points

- The paper addresses the challenge of improving AV mapping and navigation in work zones
- Current solutions struggle to adapt to temporarily changed drivable areas like work zones
- The proposed prototype incorporates crowdsourcing vehicle trajectories into the mapping process
- A Gaussian Mixture Model (GMM) is used to construct a temporarily changed drivable area and occupancy grid map (OGM)
- Crowdsourcing trajectories significantly improve AV's ability to define drivable areas accurately
- Pure SLAM method fails to achieve compliance with driving rules in work zone scenarios
- This research contributes to advancing AV technology by leveraging crowd-sourced data for work zones
- Integration of human driving experiences enhances AVs' ability to adapt to temporarily changed environments
- Possibilities for further research on incorporating crowd-sourced data into other aspects of AV development

## Summary

This paper addresses the challenge of improving autonomous vehicle (AV) mapping and navigation in work zones by introducing crowdsourcing vehicle trajectories into the mapping process. The prevalent solutions for AV mapping, such as high definition maps (HD maps) or real-time simultaneous localization and mapping (SLAM), rely solely on onboard sensors or embedded maps and struggle to adapt to temporarily changed drivable areas like work zones. Navigating AVs in these areas heavily depends on how the vehicle defines drivable areas based on perception information. To enhance AVs' understanding of the drivable area and traffic rules in work zones, this paper proposes a prototype that incorporates crowdsourcing trajectories information into the mapping process. The authors apply a Gaussian Mixture Model (GMM) to construct a temporarily changed drivable area and occupancy grid map (OGM) based on crowdsourced trajectories. By leveraging human driving information, this method aims to overcome the difficulties in improving perception accuracy and ensuring correct interpretation of perception results faced by AVs in work zone scenarios. The proposed method is compared with SLAM without any human driving information. The results show that the introduced crowdsourcing trajectories significantly improve AV's ability to define drivable areas accurately, leading to better downstream path planning and vehicle control modules. In contrast, the pure SLAM method fails to achieve compliance with driving rules in these situations. This research contributes to advancing AV technology by leveraging crowd-sourced data to enhance mapping and navigation capabilities specifically tailored for work zones. By integrating human driving experiences into the mapping process, AVs can better adapt to temporarily changed environments and ensure safer navigation through work zones. This approach opens up possibilities for further research on incorporating crowd-sourced data into other aspects of AV development, ultimately contributing to more efficient and reliable autonomous transportation systems. The authors acknowledge that this paper presents a prototype indicating ongoing development of their proposed method; they also mention that a journal version is currently in progress suggesting potential future publications with more detailed findings and refined methodologies. Overall, this paper provides a valuable contribution to the field of AV mapping and navigation in work zones highlighting the potential benefits of integrating crowdsourcing trajectories into the mapping process.