Borovets T. Design and analysis of intelligent state observers for the motion control system of electric vehicles.

The dissertation work is dedicated to the synthesis and analysis of state vector observers for the motion control system of an electric vehicle's wheel to improve energy efficiency, mobility, and enhance the safety of the vehicle's movement. The obtained research results allow improving the accuracy of observing immeasurable quantities in electric vehicle control systems, enabling the synthesis of a full-state control system considering the mobility reserve. In the first chapter, the problems of measuring quantities in automotive systems are disclosed, and the need for using algorithms to observe immeasurable coordinates is justified. An analysis of literary sources has been presented, in which the use of both classical and intelligent observation algorithms in wheeled vehicle control systems is considered. It is demonstrated that the mathematical models of vehicles and the dynamics of wheel-road interaction are nonlinear, which significantly limits the use of linear observers. It is shown that questions regarding the efficiency of applying classical observers in off-road conditions remain unexplored, especially during wheel motion on deformed ground. The second chapter is dedicated to the study of mathematical models that describe the dynamics of vehicle motion on deformed surfaces, taking into account the differences of various road surfaces. The approach of inverse dynamics has been further developed for analyzing the mobility margin of the vehicle and for devising methods to limit the wheel torque in order to maintain the mobility index within acceptable bounds. The need to use an observer to determine the components of the mobility index in real-time has been justified. The accuracy of classical observers for estimating rotational torque and the normal reaction, which operates at the tire-surface contact under the influence of disturbances on signals from sensors, non-Gaussian noises, and in the case of imprecise initial conditions, has been analyzed. Considering the features of the wheel module model, it is shown that in the observer's operation algorithm, it is advisable to use implicit numerical integration methods with low computational costs. In the third chapter, modifications of classical observation algorithms using fuzzy logic are considered. The use of the fuzzy approach for the synthesis of the Luenberger linear observer is shown, which involves designing an observer with several gain matrices that are adjusted for different performance levels. Also, the use of fuzzy logic to adapt the classic Luenberger observer to nonlinear systems is considered. Using the theory of fuzzy sets to implement the fuzzy Luenberger observer allows for improved observation accuracy in nonlinear "electric drive-wheel" systems compared to the classic observer. The proposed modified particle filter algorithm with dynamic changes in the number of particles and a variable distribution of particles during computation allows reducing the average computation time by 53% compared to the classical approach, while ensuring high estimation accuracy. The obtained results confirmed the effectiveness of the proposed solutions. In the fourth chapter, the results of experimental studies on the modified particle filter algorithm with dynamic changes in the number of particles and variable particle distribution are presented. The algorithm was implemented on the dSPACE MicroAutobox II controller using MATLAB/Simulink. The algorithm was tested on the MTS Flat-Trac experimental setup at the National Tire Research Center in Alton, USA. The implemented observer was tested in two operating modes: (i) with a minor vertical load of 100 kg and (ii) with a significant load of 2690 kg. The estimated values of the angular velocity, obtained by the observer, provide data for calculating the tire slip and rolling radius with high precision. The results confirm the feasibility of using the fuzzy particle filter algorithm to determine the wheel-surface interaction characteristics in real-time, allowing for the synthesis of a vehicle mobility controller. Another direction of experimental research was to determine the dependencies of the wheel's electric drive coordinates of a vehicle from the surface. To achieve this goal, a four-wheel mobile platform with individual wheel drives was developed. It was found that the movement of a wheeled vehicle on different surfaces causes different dynamics of the states of the wheel's electric drive system. For movement on asphalt, higher angular wheel velocities and lower current values are typical compared to movement on a grassy surface. Movement on an inclined surface results in higher motor driver currents. At the same time, the difference between the current curves for different terrains remains evident. These experimental results can serve as the basis for the development of intelligent surface estimators, allowing for the formulation of control effects taking into account the driving surface.