State and Parameter Estimation of Vehicle-Trailer Systems

Vehicle-trailer systems have different unstable modes that should be considered in their stability control, including trailer snaking, jack-knifing, and roll-over. In general, vehicle control systems require vehicle parameters and states, including geometric parameters, mass, tire forces, and side slip angles which some are not constant or can be measured economically. In a vehicle-trailer system, the trailer states and parameters such as articulation angle, trailer geometric parameters, trailer mass, trailer tire forces, and yaw rate need to be measured or identified/estimated, in addition to the unknown vehicle states/parameters. The trailer states and parameters can be measured by sensors such as Inertial Measurement Unit (IMU), wheel torque sensors, and force measurement units. However, most of these sensors are not commercially viable to be used in a vehicle or trailer due to significant extra costs. 
\n
\nEstimation algorithms are the other tools to identify the parameters and states of the system without imposing extra costs. Accurate state and parameter estimators are needed for the development and implementation of a stability control system for a vehicle-trailer system. The main purpose of this research is to design real-time state and parameter estimation algorithms for vehicle-trailer systems. 
\n
\n
\nCorrespondingly, a comprehensive overview of different model-based and non-model-based techniques/algorithms used for estimating vehicle-trailer states and parameters are provided. The vehicle-trailer system equations of motion are then presented and based on the presented vehicle-trailer model, the possibility of the trailer states and parameters estimation are investigated for different possible vehicle-trailer on-board sensor settings.
\n
\n
\nTwo different methods are proposed to estimate trailer mass for arbitrary vehicle-trailer configurations: model-based and Machine Learning (ML). The stability of the model-based estimation algorithm is analyzed, establishing the convergence of the estimation error to zero. In the proposed ML-based approach, a deep neural network is designed to estimate trailer mass. The inputs of the ML-based method are selected based on the vehicle-trailer model and are normalized by the vehicle mass, tire sizes, and geometry so that retraining of the network is not needed for different towing vehicles. The simulation and experimental results demonstrate that the trailer mass can be estimated with with acceptable computational costs. 
\n
\n
\nIn this thesis, ultrasonic sensors along with kinematics and dynamics equations of a towing vehicle are used to develop three approaches for hitch angle estimation. The first approach is based on direct calculation of hitch angle using certain a priori geometric information and distance measurements of four Ultra sonic sensors. As the second and third approaches, kinematic and dynamic models of the vehicle-trailer system are used to develop least-square and Kalman filter based recursive hitch angle estimations. A more reliable hitch angle estimation scheme is then proposed as the integration of the algorithms developed following each of the three approaches via a switching data fusion logic. It is shown that the proposed integrated hitch angle estimation scheme can be used for any ball type trailer with a flat or symmetric V-nose frontal face without any priori information on the trailer parameters.
\n
\n
\n
\nAdditionally, a new approach in estimating the lateral tire forces and hitch-forces of a vehicle-trailer system is introduced. It is shown that the proposed hitch-force estimation is independent of trailer mass and geometry. The designed lateral tire forces and hitch-force estimation algorithms can be used for any ball type trailer without any priori information on the trailer parameters. A vehicle-trailer model is proposed to design an observer for the estimation of the hitch-forces and lateral tire forces. Simulations studies in CarSim along with experimental tests are used to validate the presented method to confirm the accuracy of the developed observer. 
\n
\nMoreover, using the vehicle-trailer lateral dynamics along with the LuGre tire model, an estimation system for the lateral velocity of a vehicle-trailer is proposed. It is shown that the proposed estimation is robust to the road conditions. An affine quadratic stability approach is used to analyze the stability of the proposed estimation. The test results confirm the accuracy of the developed estimation and convergence of the vehicle-trailer lateral velocity estimation to the actual value.
\n
\n
\nModel-based and ML-based estimators are developed for estimating road angles for arbitrary vehicle-trailer configurations. The estimators are shown to be independent from road friction conditions. The model-based method employs unknown input observers on the vehicle-trailer roll and pitch dynamic models. In the proposed ML-based estimator, a recurrent neural network with Long-short-term-memory gates is designed to estimate the road angles. The inputs to the ML-based method are normalized by the vehicle wheel-base, mass, and CG height to make it applicable to any towing vehicle with the need of retraining. The simulation and experimental results justify the convergence of the road angle estimation error.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.