In this paper the neural network-based controller is designed for motion control of a mobile robot. This paper treats the problems of trajectory following and posture stabilization of the mobile robot with nonholonomic constraints. For this purpose the recurrent neural network with one hidden layer is used. It learns relationship between linear velocities and error positions of the mobile robot. This neural network is trained on-line using the backpropagation optimization algorithm with an adaptive learning rate. The optimization algorithm is performed at each sample time to compute the optimal control inputs. The performance of the proposed system is investigated using a kinematic model of the mobile robot. Keywords—Mobile robot, kinematic model, neural network, motion control, adaptive learning rate.

The problem of motion planning and control of mobile robots has attracted the interest of researchers in view of its theoretical challenges because of their obvious relevance in applications. From a control viewpoint, the peculiar nature of nonholonomic kinematics and dynamic complexity of the mobile robot makes that feedback stabilization at a given posture cannot be achieved via smooth time-invariant control (Oriolo et al., 2002). This indicates that the problem is truly nonlinear; linear control is ineffective, and innovative design techniques are needed. In recent years, a lot of interest has been devoted to the stabilization and tracking of mobile robots. In the field of mobile robotics, it is an accepted practice to work with dynamical models to obtain stable motion control laws for trajectory following or goal reaching (Fierro & Lewis, 1997). In the case of control of a dynamic model of mobile robots authors usually used linear and angular velocities of the robot (Fierro & Lewis, 1997; Fukao et al., 2000) or torques (Rajagopalan & Barakat , 1997; Topalov et al., 1998) as an input control vector. The central problem in this paper is reduction of control torques during the reference position tracking. In the case of dynamic mobile robot model, the position control law ought to be nonlinear in order to ensure the stability of the error that is its convergence to zero (Oriollo et al., 2002). The most authors solved the problem of mobile robot stability using nonlinear backstepping algorithm (Tanner & Kyriakopoulos, 2003) with constant parameters (Fierro & Lewis, 1997), or with the known functions (Oriollo et al., 2002). In (Tanner & Kyriakopoulos, 2003) a combined kinematic/torque controller law is developed using backstepping algorithm and stability is guaranteed by Lyapunov theory. In (Oriollo et al., 2002) method for solving trajectory tracking as well as posture stabilization problems, based on the unifying framework of dynamic feedback linearization was presented. The objective of this chapter is to present advanced nonlinear control methods for solving trajectory tracking as well as convergence of stability conditions. For these purposes we developed a backstepping (Velagic et al., 2006) and fuzzy logic position controllers (Lacevic, et al., 2007). It is important to note that optimal parameters of both controllers are adjusted using genetic algorithms. The novelty of this evolutionary approach lies in automatic obtaining of suboptimal set of control parameters which differs from standard manual adjustment presented in (Hu & Yang, 2001; Oriolo et al., 2002). The considered motion control system of the mobile robot has two levels. The lower level subsystem deals with the

This paper proposes a procedure for parametric identification of plants having multiple sources of the pure time delays and internal feedbacks, which are extremely complicated for the identification. The procedure is based on a genetic algorithm applied to the samples of the frequency response on the plant.

Abstract This paper deals with the Cross-Entropy method application in the control theory. The method is a the combinatorial optimization technique that is mostly used in the networks theory and could be used in deterministic optimization problems as well. The paper shows the possibiliy of the Cross-Entropy usage in the control parameter tuning. Similar to genetics algorithms, this method minimizes a given performance function in order to find optimal parameters. The appropirate conclusions about relability and the convergence rate of the method were experimentaly supported. The first two experiments used PI controller with different performance functions, where the optimal controller values have been obtained through the simulation. The third experiment used LQR controller to control a complex system. The tunining of four parameters of a LQR control matrix and obtained values were compared with the ones generated by LQR algorithm.

Inamodern vehicle systems oneofthemaingoals to Dmy achieve isdriver 'ssafety, andmanysophisticated systems aremade DummyAccelertion forthat purpose. Vibration isolation forthevehicle seats, andatthe Sensorl_/atio same timeforthedriver, isoneofthechallenging problems. Ss SeatCushlion Parameters ofthecontroller usedfor theisolation canbetunedfor adifferent roadtypes, making theisolation better (specially forthe = .no vehicles like dampers, tractors, field machinery, bulldozers, etc.). In Seat Acceleraton splacement Sensor this paperwepropose themethod whereneural networks areused Seat Eleratzon forroadtype recognition. Themaingoalistoobtain agoodroad Sels_ recognition forthepurposeofbetter vibration damping ofa driver 'ssemiactive controllable seat.Therecognition ofaspecific P Spn;;; roadtype will bebased onthemeasurable parameters ofavehicle.Cabin Acceleration Discrete Fourier Transform ofmeasurable parameters isobtained Sedsor andusedfortheneural network learning. Thedimension ofthe Cabin Position Sensor input vector, asthemainparameter that decides thespeed ofroad recognition, isvaried. Hydral llyl le|

An osteotomy is a surgical operation whereby a bone is cut to shorten, lengthen, or change its alignment. Corrective tibial osteotomies correct non-physiological axis, thus eliminating knee and ankle joint loads. In preoperative planning of the osteotomy a preoperative drawing should be made and it requires knowledge of biomechanics and physiology of the lower limbs.

This paper develops a fuzzy logic position controller which membership functions are tuned by genetic algorithm. The main goals are to ensure both successfully velocity and position trajectories tracking between the mobile robot and the reference cart. The proposed fuzzy controller has two inputs and two outputs. The first input represents the distance between the mobile robot and the reference cart. The second input is the angle formed by the straight line defined with the orientation of the robot, and the straight line that connects the robot with the reference cart. Outputs represent linear and angular velocity inputs, respectively. The performance of the fuzzy controller is investigated through comparison with previously developed a mobile robot position controller based on backstepping control algorithm. Simulation results obtained the good quality of both position tracking and torque capabilities with the proposed fuzzy controller. Also, sufficient improvement of torques reduction is achieved in the case of fuzzy controller.

This paper deals with both introducing novel technique of calculating population diversity and analyzing the existing ones. This motivation to investigate new methods of determining population diversity lies in significant disadvantages of commonly used techniques, particularly the ones that operate in a parameter space. The problem with these methods is that they can produce inexact information about population state, e.g. indicate high diversity when it is far from being high. For the purpose of eliminating these problems, new diversity mechanisms are investigated. The main idea was to use the information that is contained in the matrix with all mutual distances between individuals. New mechanism can be employed within a standard parallel search algorithms (whether as analyzing or guiding mechanism), or in general, as a mechanism for determining how well does the finite set of points sample a compact region of space.

This paper proposes a new reactive planning algorithm for mobile robot navigation in unknown environments. The overall navigation system consists of three navigation subsystems. The lower level subsystem deals with the control of the linear and angular velocities using a multivariable PI controller described with a full matrix. The position control of the mobile robot is in the medium level, and it is a nonlinear. The nonlinear control design is implemented by a backstepping algorithm whose parameters are adjusted by a genetic algorithm. The high level subsystem uses the Fuzzy logic and Dempster-Shafer evidence theory to design the fusion of sensor data, map building and path planning tasks. The path planning algorithm is based on a modified potential field method. In this algorithm, the fuzzy rules for selecting the relevant obstacles for robot motion are introduced. Also, suitable steps are taken to pull the robot out of the local minima. A particular attention is paid to detection of the robot's trapped state and its avoidance. One of the main issues in this paper is to reduce the complexity of planning algorithms. Simulation results show a good quality of position tracking capabilities and obstacle avoidance behavior of the mobile robot.

In this paper we proposed a new stable control algorithm for mobile robot trajectory tracking. The stability conditions are guaranteed by Lyapunov theory. The control parameters of backstepping algorithm are adjusted using genetic algorithm. Some of them are represented by unknown functions which are generated by neural network. The performance of the proposed controller is investigated using a kinematic model of a nonholonomic mobile robot. The efficient position tracking performance was obtained but the velocities were very high at the start of the motion. In order to avoid this, we proposed the extension of backstepping position controller by adding a new control law, which provided lower velocity servo inputs. Simulation results show the good quality of both velocity and position tracking capabilities of a mobile robot.

In this paper, capabilities of a feed-forward neural network regarding control of the complex object are investigated. Neural controllers have been trained by a genetic algorithm with adaptive mutation and crossover probabilities. A specific model of aggressive selection operator is proposed along with one way of co-evolution of the crossover and mutation rates. Also, different mechanisms of operator adaptation were compared in sense of resulting controller performance. Finally, the measurement results, taken from the object (hydraulically driven two-joint robot arm) are presented.

The goal of this paper is to develop an approach to increase the mobile robot navigation speed in an a priori unknown environment. The overall navigation system consists of three navigation subsystems. The lower level subsystem deals with the control of the linear and angular volocities using a multivariable PI controller described with a full matrix. The position control of the mobile robot represents the medium level control, which is implemented by a backstepping algorithm whose parameters are adjusted by a genetic algorithm. The high level subsystem uses Dempster-Shafer evidence theory to design the fusion of sensor data and map building processes. Path planning algorithm is based on the modificated potential field method. In this algorithm, the fuzzy rules for selecting the relevant obstacles for robot motion are introduced. The performance of the proposed system is investigated using a dynamic model of a mobile robot with various frictions.

We have previously developed a mobile robot position controller based on backstepping control algorithm. In this paper, we propose the extension of mentioned controller with an aim to rapidly decrease the control torques needed to achieve the desired position and orientation of mobile robot. The parameters of this controller are adjusted by genetic algorithm. The same genetic algorithm was used for evolution of the control parameters of a multivariable PI velocity controller described with a full matrix. The performance of the proposed system is investigated using a dynamic model of a nonholonomic mobile robot with friction. Simulation results show the good quality of both velocity and position tracking capabilities of a mobile robot