Hossein Javaherian

Adaptive Critic Learning Techniques for Engine Torque and Air–Fuel Ratio Control

A new approach for engine calibration and control is proposed. In this paper, we present our research results on the implementation of adaptive critic designs for self-learning control of automotive engines. A class of adaptive critic designs that can be classified as (model-free) action-dependent heuristic dynamic programming is used in this research project. The goals of the present learning control design for automotive engines include improved performance, reduced emissions, and maintained optimum performance under various operating conditions. Using the data from a test vehicle with a V8 engine, we developed a neural network model of the engine and neural network controllers based on the idea of approximate dynamic programming to achieve optimal control. We have developed and simulated self-learning neural network controllers for both engine torque (TRQ) and exhaust air-fuel ratio (AFR) control. The goal of TRQ control and AFR control is to track the commanded values. For both control problems, excellent neural network controller transient performance has been achieved.

Neural network modeling and adaptive critic control of automotive fuel-injection systems

We investigate the applications of a class of adaptive critic designs that can be classified as (model-free) action-dependent heuristic dynamic programming (ADHDP). Adaptive critic designs are defined as designs that approximate dynamic programming in the general case, i.e., approximate optimal control over time in noisy, nonlinear environment. The goals of the present learning control design for automotive engines include improved performance, reduced emissions, and maintained optimum performance under various operating conditions. Using the data obtained from a test vehicle, we first develop a neural network model of the engine. The neural network controller is then designed based on the idea of approximate dynamic programming to achieve optimal control. In the simulation studies, the initial controller is trained using the neural network engine model developed rather than the actual engine. We have developed and tested self-learning neural network controllers for both engine torque and exhaust air-fuel ratio control. The goal of the engine torque control is to track the commanded torque. The objective of the air-fuel ratio control is to regulate the engine air-fuel ratio at specified set points. For both control problems, good transient performance of the neural network controller has been observed. A distinct feature of the present technique is the controllers real-time adaptation capability which allows the neural network controller to be further refined and improved in real-time vehicle operation through continuous learning and adaptation.

Neural sliding-mode control of engine torque

Abstract In this paper, we investigate the applications of neural sliding-mode control method to automotive engine control. The scheme of neural sliding-mode control is realized by two parallel neural networks. The first neural network estimates the equivalent control term and the other one generates the corrective control term. The goal of the present learning control design of automotive engines is to track the commanded torque under various operating conditions. Using the data from a test vehicle with a V8 engine, we have developed a neural network engine model and neural network controllers based on the idea of sliding-mode control to achieve optimal torque control. In simulation studies of the neural sliding-mode design method, very good transient performance and fast speed of convergence have been observed. In this process, the tedious task of parameter tuning by trial-and-error has been eliminated. Distinct features of the present technique are the controllers real-time adaptation capability based on observed real vehicle data and its rapid convergence which allow the neural network controller to be further refined and improved in real-time vehicle operation through continuous learning and adaptation.

NONLINEAR TORQUE AND AIR-TO-FUEL RATIO CONTROL OF SPARK IGNITION ENGINES USING NEURO-SLIDING MODE TECHNIQUES

This paper presents a new approach for the calibration and control of spark ignition engines using a combination of neural networks and sliding mode control technique. Two parallel neural networks are utilized to realize a neuro-sliding mode control (NSLMC) for self-learning control of automotive engines. The equivalent control and the corrective control terms are the outputs of the neural networks. Instead of using error backpropagation algorithm, the network weights of equivalent control are updated using the Levenberg-Marquardt algorithm. Moreover, a new approach is utilized to update the gain of corrective control. Both modifications of the NSLMC are aimed at improving the transient performance and speed of convergence. Using the data from a test vehicle with a V8 engine, we built neural network models for the engine torque (TRQ) and the air-to-fuel ratio (AFR) dynamics and developed NSLMC controllers to achieve tracking control. The goal of TRQ control and AFR control is to track the commanded values under various operating conditions. From simulation studies, the feasibility and efficiency of the approach are illustrated. For both control problems, excellent tracking performance has been achieved.

Application of interval observers to estimation and control of air-fuel ratio in a direct injection engine

The problem of air-to-fuel ratio regulation for a direct injection engine is addressed. A LPV model of the engine is used, for which an interval observer is designed. The interval observer is applied for the model validation and control synthesis. The results of design are confirmed by implementation.

Identification of linear parameter-varying engine models

The real-time identification and monitoring of automotive engines has posed many challenging problems. The difficulties are mainly due to the nonlinearity of the engine dynamics due to changes in the engine operating conditions. Various recent studies have demonstrated that many of the powertrain subsystems are well approximated as linear parameter-varying (LPV) systems that are described as time-invariant linear systems with feedback multiplied by operating condition dependent parameters that can be measured or otherwise obtained in real time. The LPV structure has linear dynamics at a fixed operating condition, and has been shown to incorporate much of the known governing laws of physics directly into the structure of the dynamic model. Previously available LPV system identification methods are problematic, being iterative or involving an exponentially growing number of terms that can result in low accuracy models. A recently developed subspace method [Larimore(2013a)] avoids these difficulties giving efficient solutions on larger scale problems using well understood linear time-invariant subspace methods. An added benefit is the rigorous determination of the state order of the process that can be valuable for controller implementation. The identification of engine subsystem models in LPV form has the advantages of greatly improved accuracy, greatly reduced data requirements, and dramatic abilities to extrapolate to conditions not contained in the model fitting data. Use of accurate LPV models in other fields has led to the design of global controllers having guaranteed global stability and margin with improved performance, and monitoring methods to detect changes and faults under operating conditions not previously encountered. Potential issues are significant non-linearities of some engine models that may require the use of recently developed Quasi-LPV subspace methods. Also, to achieve the potential high identification accuracy may require the use of quadruple precision computation for SVD of very large matrices, that is starting to be practical for real-time engine model identification.

State-dependent Kalman filters for robust engine control

Vehicle emissions variations impose significant challenges to the automotive industry. In these simulation studies, nonlinear estimation techniques based on state-dependent and extended Kalman filtering are developed for spark ignition engines to enhance robustness of the feedforward fuel controllers to changes in nominal system parameters and measurement errors. A model-based approach is used to derive the optimal filters. Numerical simulations indicate the superiority of estimation-based approaches to enhance robustness of in-cylinder air estimation which directly contributes to the precision of engine exhaust air-fuel ratio and, consequently the consistency of the tailpipe emissions. The results obtained are for an aggressive driving profile and are presented and discussed

A biologically inspired adaptive nonlinear control strategy for applications to powertrain control

In this paper, an adaptive nonlinear control strategy derived from a biological control system is developed and its applications to the automotive engine are presented. The biological adaptive nonlinear control strategy inspired by the functions of baroreceptor reflex is realized by a parallel controller. The controller consists of a linear controller and a nonlinear controller that interact via a reciprocal lateral inhibitory mechanism. The linear controller design is based on a PID controller, while the nonlinear controller is constructed from neural networks which are updated online. In order to provide superior control performances, the controller must be robust to external unknown disturbances, un-modeled dynamics and plant uncertainties and also be able to perform well under a wide range of operating conditions. In the linear operating region, the linear controller takes control. If the controlled process is far away from the linear regime or is disturbed by the noise, the output of linear controller may be inappropriate, and therefore the nonlinear controller is activated to compensate for the inadequacy of the linear controller in a dynamic environment and in the presence of distances and process parameter variations. These situations can be addressed by adjusting the amount of lateral inhibition and learning the characteristics of the controlled system such that desirable controller outputs are produced in any particular operating region. The novelty of the biological adaptive nonlinear control strategy is that each controller modulates the other controller via the reciprocal lateral inhibitory connections. The good transient performance, computational efficiency, real-time adaptability and superior learning ability are illustrated through extensive numerical simulations for engine torque management driven by the biological adaptive nonlinear control strategy.

Model-Based Nonlinear Multivariable Engine Control

Optimal solutions for simultaneous air and fuel control in a spark ignition engine with the electronic throttle control are investigated. In an optimization framework, the method uses the already identified nonlinear physical models of engine processes for simultaneous torque tracking and air- fuel ratio regulation at the stoichiometric value. Simple physical arguments are used to reformulate the infeasible direct optimization problem into the optimal model predictive control framework for which a solution is sought. In the reformulated optimization problem, the engine torque is directly related to the cylinder air charge so that simpler feasible solutions for real-time control are obtained. Based on the identified engine models and predictive interpretation of the driver torque demand in a throttle-by-wire control strategy, the throttle position and the amount of fuel injection at every engine event are determined. Simulation studies of the predictive control solutions over the aggressive US06 driving cycles indicate that significant improvements in the transient accuracy of the air-fuel ratio control and fast delivery of the divers torque request through more aggressive throttle actuation are possible.

Virtual air-fuel ratio sensors for engine control and diagnostics

Virtual air-fuel ratio sensors for an internal combustion engine using recurrent neural and wavelet networks have been developed. A nonlinear state-space modeling strategy is proposed for the architecture of the stated recurrent neural network which is trained using some variants of real time recurrent learning (RTRL) algorithm. A two-stage training approach is proposed for improving the accuracy of the RNN topology. Additionally, wavelets as activation functions have been employed to construct a single-layer network called wavenet. The wavenet is used to model the exhaust air-fuel ratio that has proved a more challenging task in a purely neural net-based architecture using sigmoid activation functions. The methodology has been implemented in a V8 spark ignition engine through rapid prototyping tools for the real time generalization and performance evaluation. Observations and comments are made on the test patterns used for the training. Some of the limitations of such a data driven approach are highlighted. Representative experimental results for the 8-cylinder engine test data are listed. The virtual sensor may be used for more precise average air-fuel ratio control and enhanced reliability engendered through the diagnostic capabilities of the sensor