EPE and Speed Adaptive Extended Kalman Filter for Vehicle Position and Attitude Estimation with Low Cost GNSS and IMU Sensors

High-precision position estimations of agricultural mobile robots (AMRs) are crucial for implementing control instructions. Although the global navigation satellite system (GNSS) and real-time kinematic GNSS (RTK-GNSS) provide high-precision positioning, the AMR accuracy decreases when the signals interfere with buildings or trees. An improved position estimation algorithm based on multisensor fusion and autoencoder neural network is proposed. The multisensor, RTK-GNSS, inertial-measurement-unit, and dual-rotary-encoder data are fused with Extended Kalman filter (EKF). To optimize the EKF noise matrix, the autoencoder and radial basis function (ARBF) neural network was used for modeling the state equation noise and EKF measurement equation. A multisensor AMR test platform was constructed for static experiments to estimate the circular error probability and twice-the-distance root-mean-squared criteria. Dynamic experiments were conducted on road, grass, and field environments. To validate the robustness of the proposed algorithm, abnormal working conditions of the sensors were tested on the road. The results showed that the positioning estimation accuracy was improved compared to the RTK-GNSS in all three environments. When the RTK-GNSS signal experienced interference or rotary encoders failed, the system could still improve the position estimation accuracy. The proposed system and optimization algorithm are thus significant for improving AMR position prediction performance.

Section: Predictmentioning

confidence: 99%

Improved Position Estimation Algorithm of Agricultural Mobile Robots Based on Multisensor Fusion and Autoencoder Neural Network

Gao

Lee

Jeon

et al. 2022

“…The stiffness coefficients C α f and C α r are obtained through the procedure described in Section 3.1.2. The process noise covariance matrix (Q n ) in a vehicle model is suggested to be calculated as the propagation of each value per time step [11], in this sense, gathering the standard deviation of parameters from the test vehicle circulating in normal conditions helps to determine Q n .…”

Section: Adaptive Unscented Kalman Filtermentioning

confidence: 99%

“…To test the proposed approaches, the following parameter values have been applied. The noise covariances for the UKF have been calculated as suggested by [11], the process noise covariance matrix (Q n ) is defined assuming the standard deviation of parameters from the test vehicle circulating in normal conditions helps to determine Q n . The measurement noise covariance matrix (R n ) is selected by taking into account the accuracy of commercially available acquisition devices.…”

Section: Parametersmentioning

confidence: 99%

“…A detailed description of the most popular methods and techniques for performing data fusion is presented in [9], where the author concludes that the appropriate technique to be implemented depends on the type of problem. In the automotive field, the Bayesian approach, extended and unscented Kalman filters (UKF) are mostly used [10][11][12]. However, these techniques depend mainly on the information directly from on-board sensors without any fall-back strategy.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Fail-Operational Control Architecture Approach and Dead-Reckoning Strategy in Case of Positioning Failures

Matute

Pérez

2020

Presently, in the event of a failure in Automated Driving Systems, control architectures rely on hardware redundancies over software solutions to assure reliability or wait for human interaction in takeover requests to achieve a minimal risk condition. As user confidence and final acceptance of this novel technology are strongly related to enabling safe states, automated fall-back strategies must be assured as a response to failures while the system is performing a dynamic driving task. In this work, a fail-operational control architecture approach and dead-reckoning strategy in case of positioning failures are developed and presented. A fail-operational system is capable of detecting failures in the last available positioning source, warning the decision stage to set up a fall-back strategy and planning a new trajectory in real time. The surrounding objects and road borders are considered during the vehicle motion control after failure, to avoid collisions and lane-keeping purposes. A case study based on a realistic urban scenario is simulated for testing and system verification. It shows that the proposed approach always bears in mind both the passenger’s safety and comfort during the fall-back maneuvering execution.

“…As pointed out in Reference [48], the KF provides a suitable and easy to implement solution for a variety of state estimation problem statements. Originally, the applicability of the KF assumes that the underlying process information and given measurements can be formulated with linear equations:xk=Φk−1xk−1+wk−1 zk=Hkxk+vk.…”

Section: Fundamentalsmentioning

confidence: 99%

An Empirical Study on V2X Enhanced Low-Cost GNSS Cooperative Positioning in Urban Environments

Schwarzbach

Michler

Tauscher

et al. 2019

High-precision and lane selective position estimation is of fundamental importance for prospective advanced driver assistance systems (ADAS) and automated driving functions, as well as for traffic information and management processes in intelligent transportation systems (ITS). User and vehicle positioning is usually based on Global Navigation Satellite System (GNSS), which, as stand-alone positioning, does not meet the necessary requirements in terms of accuracy. Furthermore, the rise of connected driving offers various possibilities to enhance GNSS positioning by applying cooperative positioning (CP) methods. Utilizing only low-cost sensors, especially in urban environments, GNSS CP faces several demanding challenges. Therefore, this contribution presents an empirical study on how Vehicle-to-Everything (V2X) technologies can aid GNSS position estimation in urban environments, with the focus being solely on positioning performance instead of multi-sensor data fusion. The performance of CP utilizing common positioning approaches as well as CP integration in state-of-the-art Vehicular Ad-hoc Networks (VANET) is displayed and discussed. Additionally, a measurement campaign, providing a representational foundation for validating multiple CP methods using only consumer level and low-cost GNSS receivers, as well as commercially available IEEE 802.11p V2X communication modules in a typical urban environment is presented. Evaluating the algorithm’s performance, it is shown that CP approaches are less accurate compared to single positioning in the given environment. In order to investigate error influences, a skyview modelling seeking to identify non-line-of-sight (NLoS) effects using a 3D building model was performed. We found the position estimates to be less accurate in areas which are affected by NLoS effects such as multipath reception. Due to covariance propagation, the accuracy of CP approaches is decreased, calling for strategies for multipath detection and mitigation. In summary, this contribution will provide insights on integration, implementation strategies and accuracy performances, as well as drawbacks for local area, low-cost GNSS CP in urban environments.