Search

This paper proposes a transfer learning approach to recalibrate our previously developed Wheel Odometry Neural Network (WhONet) for vehicle positioning in environments where Global Navigation Satellite Systems (GNSS) are unavailable. The WhONet has been shown to possess the capability to learn the uncertainties in the wheel speed measurements needed for correction and accurate positioning of vehicles. These uncertainties may be manifested as tyre pressure changes from driving on muddy and uneven terrains or wheel slips. However, a common cause for concern for data-driven approaches, such as the WhONet model, is usually the inability to generalise the models to a new vehicle. In scenarios where machine learning models are trained in a specific domain but deployed in another domain, the model's performance degrades. In real-life scenarios, several factors are influential to this degradation, from changes to the dynamics of the vehicle to new pattern distributions of the sensor's noise, and bias will make the test sensor data vary from training data. Therefore, the challenge is to explore techniques that allow the trained machine learning models to spontaneously adjust to new vehicle domains. As such, we propose the Recalibrated-Wheel Odometry neural Network (R-WhONet), that adapts the WhONet model from its source domain (a vehicle and environment on which the model is initially trained) to the target domain (a new vehicle on which the trained model is to be deployed). Through a performance evaluation on several GNSS outage scenarios - short-term complex driving scenarios, and on longer-term GNSS outage scenarios. We demonstrate that a model trained in the source domain does not generalise well to a new vehicle in the target domain. However, we show that our new proposed framework improves the generalisation of the WhONet model to new vehicles in the target domains by up to 32%., Comment: arXiv admin note: text overlap with arXiv:2104.02581

Low-cost inertial navigation sensors (INS) can be exploited for a reliable tracking solution for autonomous vehicles. However, position errors grow exponentially due to noises in the measurements. Several deep learning techniques have been investigated to mitigate the errors for a better navigation solution [1-10]. However, these studies have involved the use of different datasets not made publicly available. The lack of a robust benchmark dataset has thus hindered the advancement in the research, comparison and adoption of deep learning techniques for vehicle positioning based on inertial navigation. In order to facilitate the benchmarking, fast development and evaluation of positioning algorithms, we therefore present the first of its kind large-scale and information-rich inertial and odometry focused public dataset called IO-VNBD (Inertial Odometry Vehicle Navigation Benchmark Dataset).The vehicle tracking dataset was recorded using a research vehicle equipped with ego-motion sensors on public roads in the United Kingdom, Nigeria, and France. The sensors include a GPS receiver, inertial navigation sensors, wheel-speed sensors amongst other sensors found on the car as well as the inertial navigation sensors and GPS receiver in an android smart phone sampling at 10HZ. A diverse number of scenarios and vehicle dynamics are captured such as traffic, round-abouts, hard-braking etc. on different road types (country roads, motorways etc.) with varying driving patterns. The dataset consists of a total driving time of about 40 hours over 1,300km for the vehicle extracted data and about 58 hours over 4,400 km for the smartphone recorded data. We hope that this dataset will prove valuable in furthering research on the correlation between vehicle dynamics and its displacement as well as other related studies