Adaptive Fuzzy Logic Controller for Nonlinear Vehicle Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current self-driving vehicle technologies, relying on sensor fusion and artificial intelligence, struggle to accurately perceive and respond to surroundings, especially in nonlinear dynamics and with actuator uncertainties, limiting their ability to provide stable and adaptive control.
Innovation Solution
The apparatus employs an adaptive fuzzy logic controller with coarse and fine tuning assemblies, incorporating fuzzy descriptions and neural networks to model driver behavior, environmental factors, and actuator uncertainties, along with a non-linear dynamic linearized regression controller and cascaded diophantine frequency synthesis for real-time stabilization and optimal control sequence computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If linear model predictive control (MPC) is used for vehicle autonomous driving, then the control scheme is simple and computationally efficient, but it cannot adequately handle nonlinear plant dynamics and actuator uncertainties
Solution Approach 1:
The patent transforms the fixed linear model parameters into adaptive parameters that change based on operating conditions. The neural network dynamically adjusts the plant model parameters to reflect current nonlinear dynamics, while the H-infinity controller adapts its control strategy based on updated system characteristics, resolving the contradiction between model simplicity and nonlinear handling capability
Solution Approach 2:
The patent introduces dynamic adaptation mechanisms where the neural network continuously learns from new data and updates the plant model in real-time. This dynamic parameter adjustment allows the control system to adapt to changing nonlinear dynamics and actuator uncertainties, maintaining reliability without requiring a completely complex nonlinear controller
2Adaptability or versatility
If sensor-based solutions and artificial intelligence are used to perceive surroundings, then the system can process multiple sensor inputs, but it struggles to accurately understand and respond to the vehicle's environment
Solution Approach 1:
The patent implements a feedback loop where sensor measurements continuously inform the neural network, which updates the plant model. This closed-loop system refines environmental perception by constantly comparing predicted states with actual sensor readings, improving measurement precision while maintaining multi-sensor integration capability
Solution Approach 2:
The neural network acts as an intermediary between raw sensor data and the control system. It processes and interprets sensor inputs to create an accurate internal model of the environment and plant dynamics, bridging the gap between multiple sensor inputs and meaningful environmental understanding
3Reliability
If adaptive control with actuator uncertainty modeling is implemented, then the controller can handle input faults and mechanical uncertainties, but the controller performance degrades significantly without proper adaptation
Solution Approach 1:
The patent performs preliminary adaptation by pre-training the neural network on actuator uncertainty models and fault scenarios. This preliminary learning enables the controller to anticipate and compensate for actuator uncertainties before they affect performance, maintaining both robustness and high productivity during actual operation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Apparatus for controlling a land vehicle which is self-driving or partially self-driving, which apparatus comprises a coarse tuning assembly (1, 2, 3) and a fine tuning assembly (4), the coarse tuning assembly (1, 2, 3) being such that It comprises: a. a sensor interface (1) which measures kinematic parameters including speed and braking, b. fuzzy descriptions to model guidance, navigation and control of the vehicle, the fuzzy descriptions including: (i) driver behaviour and driving dynamics, (ii) uncertainties due to the environment including weather, road conditions and traffic, and (iii) input faults including mechanical and electrical parts, and c. an adaptive fuzzy logic controller (3) for nonlinear MIMO systems (2) with subsystems which comprise fuzzification, inference, and output processing, which comprise both type reduction and defuzzification, and which provide stability of a resulting closed-loop system, the adaptive fuzzy logic controller (3) including: (i) inference engine identifying relationships using a rule base and outputs as 'fuzzy sets' to a type reducer, and (ii) output control demands including torque actuators to the fuzzyfier 'fuzzyfiying' the signal, and the fine tuning assembly (4) being such that it comprises: a. inputs from the coarse tuning assembly (1, 2, 3), b. precognition horizons determining how many future samples the objective function considers for minimization and the length of the control sequence computed, c. a linearized MIMO regression model extracted from the adaptive fuzzy logic controller (3) at each time step providing the 'fine' tuning parameters, and d. a non-linear dynamic linearized regression controller (4a) providing: (I) a crisp output signal feeding into APACC synthesis (4b) computing the optimal future vehicle guidance, navigation and control sequence, and (II) reduced set output and APACC synthesis (4b) feeding into the APACC linear logic system.