مجله کنترل
سال ششم شماره 3 (پاییز 1391)

Applying supervised learning in robot navigation encounters serious challenges such as inconsistence and noisy data, difficulty to gathering training data, and high error in training data. Reinforcement Learning (RL) capabilities such as lack of need to training data, training using only a scalar evaluation of efficiency and high degree of exploration have encourage researcher to use it in robot navigation problem. However, RL algorithms are time consuming also have high failure rate in the training phase. Here, a novel idea for utilizing advantages of both above supervised and reinforcement learning algorithms is proposed. A zero order Takagi-Sugeno (T-S) fuzzy controller with some candidate actions for each rule is considered as robot controller. The aim of training is to find appropriate action for each rule. This structure is compatible with Fuzzy Sarsa Learning (FSL) which is used as a continuous RL algorithm. In the first step, the robot is moved in the environment by a supervisor and the training data is gathered. As a hard tuning, the training data is used for initializing the value of each candidate action in the fuzzy rules. Afterwards, FSL fine-tunes the parameters of conclusion parts of the fuzzy controller online. The simulation results in KiKS simulator show that the proposed approach significantly improves the learning time, the number of failures, and the quality of the robot motion.

Antilock brake system (ABS) is an active automobile safety system to achieve the maximum negative acceleration during the braking process. Also, ABS increases the automobile stability and reduces the stopping distance. Modeling the brake system and particularly ABS are very important, due to that ABS is the foundation of other advanced automobile control systems like EBD, ESP and ACC. In this paper, we introduce a detailed model of hydraulic ABS. Also, in this paper, we evaluate and validate the presented modeling, by comparing the results of simulations to experimental tests.

This paper presents a new algorithmic method to design PI controller for a class of nonlinear systems whose state space description is in the form of polynomial functions. Design procedure is taken place based on certain or uncertain nonlinear model of system and sum of squares optimization. A so called density function is employed to formulate the design problem into a convex optimization program of sum of squares optimization form. Robustness of the design is guaranteed by taking parametric uncertainty into account with an approach similar to that of generalized S-Procedure. Validity and applicability of the proposed method is certified with numerical simulation. This paper, besides presenting an innovated PI control design which is not based on local linearization and works globally, announces a new approach in formulating parametric uncertainty in nonlinear systems. Derived stability conditions do not suffer from any drawbacks seen in previous results, such as depending on a linearized model or a stable model and it can overcome most control difficulties. Furthermore, employing sum of squares techniques makes it possible to drive stability conditions with least conservatism and directly derive stability of nonlinear system.

In this paper, a method for numerical solution of differential-algebraic equations (DAEs) in Hessenberg form is presented. In this method, a sliding surface proportional to systems index is defined that generates a complete equation to calculate algebraic variables. Since the sliding surface is stable, convergence of the distance from manifold constraint in DAE is satisfied. Finally, the proposed method is applied for some linear and nonlinear index-3 systems. Numerical solutions confirm the accuracy of the proposed technique.

In this paper, a novel method is introduced for optimum energy harvest from windfarms. In the proposed method, wind farm is modeled by fuzzy-logic and the model is updated using a combination of wind parameters history and wind’s spatial information. Utilizing this model, the parameters for the wind blowing through each turbine in the wind farm is estimated. To evaluate the performance of the proposed method two practical wind types are simulated. In the first scenario, the wind maintains low turbulence and its parameters change slowly while in the second scenario the wind demonstrates high turbulences and its parameters undergo sudden shifts. Simulation results for the proposed method are obtained in both scenarios. For the first scenario, thecomparison reveals that the proposed method improves the accuracy of wind speed estimation and the monotonousness of the obtained electrical voltage by 5.3% and 0.52 volts respectively compared to existing methods. These improvements reach 17.1% and 12.7 volts in the presence of high turbulence winds in the second scenario. Based on these corroborating simulations, it is concluded that the proposed method provides a more accurate estimate of wind parameters for the wind blowing through the wind farm.

In this paper, a sliding mode controller for time-delay Markovian jump systems with partly unknown transition probability matrix in presence of disturbance is designed. The proposed method is quite general and includes both systems with completely known and completely unknown transition probability rates. At first, sufficient conditions for existence of linear switching surface are obtained in terms of linear matrix inequalities (LMIs) that guarantee the stochastic stability of sliding mode dynamics. Then, a sliding mode controller is designed such that the closed-loop system’s state trajectories reach the desired sliding surface in a finite time and maintain there for all subsequent times. As a result, the stochastic stability of closed-loop system is guaranteed by applying a specifically designed control law. All of the conditions are presented in terms of linear matrix inequalities and can be simply solved by means of numerical software tools. Finally, anumerical example is given to demonstrate the validity and effectiveness of the proposed method.

The motions which are achievable by Limit Cycle Walkers are energetically efficient and natural looking. But their capability in external disturbance rejection is still an unexplored field of study in comparison with ZMP based walkers. In this paper a planar, under-actuated and hybrid Limit Cycle Walker with seven degrees of freedom (DOF) is considered. During walking, it is assumed that the robot is subjected to an impulsive external disturbance. First, some maps have been obtained to relate the states of the system just before and just after the impact events. Then, the control is done based on the determination of holonomic constraints for the event-based feedback controller. Several simulations have been done considering disturbances exerted during walking. The results showed the performance of the method in recovery of disturbances occurring in the sagittal plane in both anterior and posterior directions. Moreover, the results showed that the simulated motions can be characterized in terms of strategies observed in human for balance recovery against perturbations during walking.