Radar-Based Multisensor Fusion for Uninterrupted Reliable Positioning in GNSS-Denied Environments

Abstract

Multi-sensor integration is necessary to provide high-precision navigation solutions for autonomous vehicles. Land-vehicles often rely on global navigation satellite systems (GNSS) to acquire its position. However, there are some environments where GNSS signals are unavailable, such as indoor-parking garages, tunnels, and under bridges. Additional sensors are required to allow reliable positioning regardless of location. The vehicle’s on-board low cost inertial sensors (accelerometers and gyroscopes) are used for positioning in GNSS-denied environments. Despite their fidelity in short-term usage, inertial navigation systems (INS) are susceptible to drifts in their positioning solution due to the inherent inertial sensor errors, causing positioning errors over time in prolonged scenarios, such as indoor-parking garages. Modern land-vehicles can be equipped with a diversified set of perception systems (e.g. LiDAR, cameras) to provide information about the surrounding environment. These systems are well-studied in literature to provide accurate positioning. Yet, the performance of these systems may degrade in weather conditions such as heavy snow or rain, and may fail in degraded vision environments. The proposed research addresses some of the limitations of current positioning technologies for land vehicles by integrating low-cost on-board motion sensors with the all-weather electronic scanning radar (ESR) systems presently used in adaptive cruise control. This research employs a method of estimating the vehicle position based on multiple ESR systems. Integration with the on-board motion sensors guarantees continuous positioning estimation for uninterrupted navigation. The multi-sensor system integration utilizes extended Kalman filter (EKF) with a unique dynamic tuning approach for reliable and uninterrupted positioning. The proposed solution was examined in real GNSS-denied scenario of an indoor parking. This research has reached an uninterrupted self-contained EKF-based multi-sensor positioning system providing less than 2m error 90% of the time in a GNSS-denied environment for up to four minutes.