Vehicle Localization Using Passive and Active Infrastructure Enable Autonomy in GPS Denied Environments

Abstract

Autonomous vehicles (AV) have the potential to provide dividends for society in areas such as alleviating congestion, reducing emissions, and increasing safety. Fundamentally, an AV makes decisions based on its surroundings and where it is located in the environment which makes localization the most critical function of an AV. The majority of autonomous transportation solutions have become highly reliant on the availability of a Global Positioning System (GPS) for localization. However, GPS is susceptible to poor signal reception in areas such as urban canyons as well as its vulnerability to malicious attacks. This research approaches solving the issue of GPS reliance by two methods — passive and active infrastructure enabled autonomy (IEA). The basis of IEA is to enable AV functions by offloading some of the computational cost to the infrastructure. The passive IEA method leverages SmartCodes (SC) in the environment where each SC has GPS information embedded in its digital signature that must be read with a camera. A radar and on-board vehicle sensors were also used and fed into an Extended Kalman Filter to localize the vehicle. The active IEA method leverages a multi-smart-sensor pack that is outfitted with road-side units that hold cameras and a radar that provide localization information to the vehicle. All data is fed into a multi-target tracking version of the linear Kalman Filter. Each method was experimentally tested and both passive and active IEA were capable of localizing a vehicle globally. The passive IEA method was able to localize a vehicle with longitudinal and lateral root-mean-square errors (RMSE) less than 0.39m and 0.18m, respectively. The active IEA method localized an average longitudinal RMSE of 4.54m and an average lateral RMSE error of 0.46m.