Four-Wheeled Dead-Reckoning Model Calibration using RTS Smoothing

Abstract

Localization is one of the main challenges to be addressed to develop autonomous vehicles able to perform complex maneuvers on roads opened to public traffic. Having an accurate dead-reckoning system is an essential step to reach this objective. This paper presents a dead-reckoning model for carlike vehicles that performs the data fusion of complementary and redundant sensors: wheel encoders, yaw rate gyro and steering wheel measurements. In order to get an accurate deadreckoning system with a drift reduced to the minimum, the parameters have to be well calibrated and the procedure has to be simple and efficient. We present a method able to accurately calibrate the parameters without knowing the ground truth by using a Rauch-Tung-Striebel smoothing scheme which enables to obtain state estimates as close to the ground truth as possible. The smoothed estimates are then used within a optimization process to calibrate the model parameters. The method has been tested using data recorded from an experimental vehicle on public roads. The results show a significant diminution of the dead-reckoning drift compared to a commonly used calibration method. We evaluate finally the average distance a vehicle can navigate without exteroceptive sensors by using the proposed four-wheeled dead reckoning system.

Improved Data Association Using Buffered Pose Adjustment for Map-Aided Localization

“…The measurements used for the DR estimation of the vehicle are a yaw rate gyro, 4 wheel encoders and the steering wheel angle. The observation models used for these sensors follows Ackerman geometry and are described in-depth in [20].…”

LiDAR Observations by Motion Compensation and Scan Accumulation

“…In the absence of accurate DR sensors, other approaches could use ICP or derivatives to register these scans prior to ground and road filtering. However, as our DR model, described in [17], is fairly stable over short periods of time (e.g: two seconds), we had no need for such registrations.…”

Estimating the reliability of georeferenced lane markings for map-aided localization

“…Hence, it is assumed that it is null. Also, in this work, the estimation of the vehicle heading is assumed to be fairly good since the evolution model used is well calibrated [14]. Therefore, the rotational component will not be estimated.…”

“…Filtering: Now that the association between track and map marking has been done and outliers have been rejected, we can use these measurements in a Kalman filtering scheme as explained in subsection II-A (a more detailed description of the filtering scheme can be found in [14]). Here, the measurements are the values y j i of each track and the measurements a priori are their projections m y j i .…”

“…Through the vehicle CAN bus were available: four wheel encoders, a gyro and the steering wheel angle. These were used for the odometric estimation [14]. The vehicle was also equipped with a Mobileye EyeQ3 smart camera which provided up to four lane marking observations at a time.…”