Data Fusing Algorithms for Improved Localization within V2X Network Using Low Cost GPS
Jing Sun, PhD
Professor, Naval Architecture and Marine Engineering
May 2016 – April 2018
Ding Zhao, PhD
Research Fellow, U-M Transportation Research Institute
Graduate Research Assistant, Department of Naval Architecture and Marine Engineering
To improve the positioning accuracy of non-RTK GPS from meters to centimeters using V2X communication.
- Build an accurate base-map using data collected from the MTC Pillar 1 project.
- Query Safety Pilot Model Deployment and Pillar 1 data.
- Develop algorithms to improve positioning accuracy based on multiple vehicle GPS information and the base-map of the road network.
- Test the algorithms on smartphones and demonstrate their potential application to pedestrian safety in Mcity.
- Validate the approach in the Pillar 1 living laboratory.
We showed that the CMM Network has the potential to provide cooperative localization with the desired accuracy for connected vehicles as a low-cost GNSS method with in both centralized and distributed manner, and can be applied on real work traffic scenarios. The following key findings led us to this conclusion:
- A Rao-Blackwellized particle filter has been proposed for the simultaneous estimation of GNSS common biases and vehicles cooperative localization using map matching
- The impact of road configuration on the CMM localization accuracy was studied theoretically, which was then used to evaluate the CMM accuracy in the real world
- A theoretical study was presented for quantitatively evaluating effects of the road constraints on the CMM accuracy and for eventually optimization of the CMM network
- A fusion mechanism for distributed CMM was provided and evaluated the correlation between the estimation variance the MSE over the network
- An implementation of CMM on real-world traffic was presented, based on Safety-Pilot database for dynamic vehicles on roads, and considered DSRC packet loss