Autonomous Navigation in Inclement Weather Based on a Localizing Ground Penetrating Radar

Abstract

Most autonomous driving solutions require some method of localization within their environment. Typically, onboard sensors are used to localize the vehicle precisely in a previously recorded map. However, these solutions are sensitive to ambient lighting conditions such as darkness and inclement weather. Additionally, the maps can become outdated in a rapidly changing environment and require continuous updating. While LiDAR systems don't require visible light, they are sensitive to weather such as fog or snow, which can interfere with localization. In this letter, we utilize a Ground Penetrating Radar (GPR) to obtain precise vehicle localization. By mapping and localizing using features beneath the ground, we obtain features that are both stable over time, and maintain their appearance during changing ambient weather and lighting conditions. We incorporate this solution into a full-scale autonomous vehicle and evaluate the performance on over 17 km of testing data in a variety of challenging weather conditions. We find that this novel sensing modality is capable of providing precise localization for autonomous navigation without using cameras or LiDAR sensors. Index Terms-Autonomous vehicle navigation, field robots, wheeled robots, intelligent transportation systems. I. INTRODUCTIONR OBUST localization in diverse conditions is a key challenge to enable the widespread deployment of autonomous vehicles. Since relying purely on a global navigation satellite system (GNSS), such as the Global Positioning System (GPS) does not provide sufficient precision, research and industry efforts have focused primarily on utilizing cameras and laser scanners for the navigation task. These systems typically use Simultaneous Localization and Mapping (SLAM) algorithms [1] for creating and maintaining maps of the environment that allow for highly precise localization and navigation.Using vision and laser sensors as localization sources comes with its own set of challenges. For instance, localization from purely visual maps needs to account for the fact that, even in the absence of occlusions, the appearance of the environment strongly varies depending on weather, season, time of day, and a potentially changing environment [2]. While laser scanners

GROUNDED: The Localizing Ground Penetrating Radar Evaluation Dataset

“…Ground Penetrating Radars: Only a few works considered the use of ground penetrating radars in robotics such as for landmine detection [23] or for autonomous surveys [24]. Using GPRs for localization has so far been considered only in [10], [11]. Consequently, most GPR datasets are targeted at very different application domains, e.g., for research on soil structure characterization [25] or meteorology [26].…”

“…This can be challenging because LGPR data can be affected by the moisture content and temperature of the underground soil which can vary with surface weather conditions. In [11] a degradation in localization performance in rain and snow was measured, but their algorithm did not explicitly account for weather changes. The second requirement, is to build maps that can localize a vehicle while it is changing between multiple lanes.…”

“…surface-based maps. LGPR was used as the sole localization sensor in [11] to navigate an automated vehicle in clear weather, rain, and snow conditions. However, the sensor demonstrated in that work is not commercially available.…”

“…We believe the dataset will enable the research community to replicate and improve upon the current state-of-the-art, and to tackle new open problems for autonomous driving in difficult weather. For example, [11] found a degradation in localization performance when localizing in rain or snow, perhaps due to the unmodeled changes to the moisture content of the soil. Furthermore, both [10] and [11] conducted both mapping and localization in a single lane.…”

“…For example, [11] found a degradation in localization performance when localizing in rain or snow, perhaps due to the unmodeled changes to the moisture content of the soil. Furthermore, both [10] and [11] conducted both mapping and localization in a single lane. Since practical autonomous vehicles will need to change lanes, it is an important area of research to stitch together multiple lanes to form a coherent map.…”