Utilizing the Infrastructure for Ground Truth Based Localization
Abstract
This paper seeks to implement new approaches for autonomous vehicle localization through utilizing the infrastructure. In both approaches a monocular vision system is used. An infrastructure enabled autonomy approach is taken wherein a camera mounted at 10 meters on roadside units is used along with vision processing to detect and localize vehicles on the road. After, a new technology developed by 3M, Smart Codes, is utilized by using an infrared camera mounted on a vehicle with Smart Code signs along the roadside to provide localization estimates. Filtering algorithms were developed for each specific localization method and experiments were performed to test the accuracy of all these localization approaches. It was found that the infrastructure enabled autonomy approach was able to estimate the vehicle’s position to within a meter over a range of about 50 meters. The 3M Smart Code approach was also able to achieve accuracy to within a meter, but over a range of approximately 20 meters. The Smart Code approach performed more consistently and saw less variation in the accuracy of the position estimates. Both approaches saw improvements through the incorporation of filtering algorithms that were able to give better, more consistent position estimations.
Citation
Marr, Tyler Stephen (2019). Utilizing the Infrastructure for Ground Truth Based Localization. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /200700.