Antonio Moran

NONLINEAR DESIGN APPROACH TO FOUR-WHEEL-STEERING SYSTEMS USING NEURAL NETWORKS

SUMMARY Four-wheel-steering (4WS) systems have been studied and developed with remarkable success from the viewpoint of vehicle dynamics. Most of the control methods require a linearized bicycle model of the actual vehicle system which is however strongly influenced by tire nonlinearity. This paper proposes a new method to design the 4WS system taking into account the nonlinear characteristics of tires and suspensions. For this purpose integration of artificial neural network and linear control theory is introduced for the identification and control of a nonlinear vehicle model structured using a software for multi-body dynamic analysis (ADAMS). This model takes into account the nonlinear characteristics of actual vehicles with tires modeled by “magic formula“. The results of computer simulations show that the proposed nonlinear approach is efficient in improving the handling and stability of vehicles.

Identification and control of nonlinear active pneumatic suspension for railway vehicles, using neural networks

Abstract This paper analyzes the performance of neural networks for the identification and optimal control of active pneumatic suspensions of high-speed railway vehicles. It is shown that neural networks can be efficiently trained to identify the dynamics of nonlinear pneumatic suspensions, as well as being trained to work as (sub)optimal nonlinear controllers. The performance of the nonlinear suspension with the neuro-controller is compared with the performance of the suspension with an LQ controller designed after linearizing the suspension components around the equilibrium point.

Analysis of the dynamics and nonlinear control of under-actuated brachiation robots

This paper analyzes the dynamics and control of underactuated nonlinear mechanical systems which are systems with fewer actuators than number of degrees-of-freedom. Two nonlinear control methods are proposed and compared for the control of underactuated systems and applied to the control of a three-link brachiation robot. Although controllers designed by both methods are able to drive the brachiation robot to the desired position, the motion characteristics and described trajectories are different for both controllers. One of the control methods based on a separation of the state variables in active and passive variables, presents better control performance and a trajectory more uniform and symmetric than the other control method based on standard nonlinear servo-control with constant desired position.

Design of controllers and observers for mimo nonlinear systems

This paper proposes new methods for design ing observers and controllers for nonlinear systems. The design procedure is reduced to find proper ob server and controller gains. The validity of the proposed methods is verified by designing output feed back controllers for the positioning control of two link robot arms.

Identification and Control of Nonlinear Active Pneumatic Suspension for Railway Vehicles Using Neural Networks

Abstract This paper analyzes the performance of neural networks for identification and optimal control of active pneumatic suspensions of high speed railway vehicles. It is shown that neural networks can be efficiently trained to identify the dynamics of nonlinear pneumatic suspensions as well as trained to work as optimal nonlinear controllers. The performance of the nonlinear suspension with neuro-controller is compared with the performance of the suspension with LQ controller designed after linearizing the suspension components around the equilibrium point.

Integration of Linear Systems and Neural Networks for Identification and Output Feedback Control of Nonlinear Systems

Abstract This paper analyzes the integration of linear systems and neural networks for the identification, state estimation and output feedback control of weakly nonlinear systems. Considering previous knowledge about the system given by approximated linear state-space models, linear observers and linear controllers; training algorithms for neuro-identification, state neuroestimation and neuro-control were derived considering the dynamics of the nonlinear system. It was found that the integrated linear-neuro model can identify the dynamics of the system much more accurately than a purely linear model or a purely neuro model. It was also found that the state estimation and control performance of the system with integrated linear-neuro output feedback control is better than the system with linear control or neuro-control.

Mixed H/sub 2//H//sub/spl infin// output feedback controller

This paper presents the solution to the output feedback mixed H/sub 2/H//sub/spl infin// control problem via the solution of an associated Nash game. Consider ing a two-player nonzero sum game, two perfor mance criteria are used, one reflecting an H//spl/sub infin// constraint and the another reflecting an H/sub 2/ optimal ity requirement. Assuming that the mixed H/sub 2/H//sub/spl infin// controller has the separation structure of H/sub 2/ and H//sub/spl infin// controllers, the feedback and observer gains of the mixed H/sub 2/H//sub/spl infin// controller are determined by solving five coupled Riccati equations.

Integration of linear systems and neural networks for identification and control of nonlinear systems

This paper analyzes the integration of linear sys tems and neural networks for the identification and optimal control of weakly nonlinear systems. Considering previous knowledge about the system given by approximated linear state-space equation mod els and linear controllers, training algorithms for identification and control were derived considering the dynamics of the nonlinear system. It was found that the integrated linear-neuro model can identify the dynamics of the system much more accurately than a purely linear model or a purely neuro model. It was also found that the vibration isolation pefor mance of the system with integrated linear-neuro control is much better than the system with linear control or neuro-control.

Shortest Trajectory Control of Autonomous Mobile Robots Using Nonlinear Observers

Abstract This paper deals with two problems related to autonomous mobile robots. 1 lie first problem is to determine the control strategy so that the robot describes the shortest trajectory linking an arbitrary position inside a working area and a path to be followed. This problem is solved by considering an equivalent minimum-time control problem and the control switching functions for obtaining the shortest trajectory has been determined in a general form. The second problem is the navigation problem and consists in designing a nonlinear observer for the dynamic estimation of the variables required to implement the shortest-trajectory control. A Luenberger-like observer for MIMO nonlinear systems is proposed and analyzed. Theoretical and experimental analyses show that the observer converges fastly enough so that path tracking can be efficiently implemented.

Integration of Neural Network and Linear Control Theory Applied to Four-Wheel-Steering Vehicles.

In order to improve handling and stability of automobiles, four-wheel-steering (4WS) systems have been studied and developed with remarkable success. Most of the control methods require a linearized two-wheel model of the actual vehicle system which is however strongly influenced by tire nonlinearity especially in critical situations such as emergency maneuvering or collision avoidance on slippery road surface. We propose a new method of designing the four-wheel-steering system, taking into account the nonlinear characteristics of tires and suspensions. For this purpose a new method using an artificial neural network and linear control theory is analyzed and applied to the identification and control of a nonlinear vehicle model structured using software for multibody dynamics (ADAMS). This model takes into account the nonlinear characteristics of actual vehicles with tires which are modeled by the Magic Formula. The results of computer simulation show that the proposed method using neural networks can be efficiently applied to improve the performance of the vehicle.