Asad Ullah Awan

Application of Echo-State Networks to the Position Control of Shape-Memory Alloys

Abstract Shape-memory alloys (SMA) have the ability to generate strain in response to temperature change. However, the relationship between applied temperature and crystalline phase is hysteretic. Obtaining an explicit representation of this highly nonlinear phenomenon for the purpose of control consumes time and effort. Also, the identification process is subject to uncertainties, and furthermore, the dynamic properties of SMAs may change during its lifetime which reduces the reliability of the identification. With this in mind, we employ an online learning control framework for the position control of an SMA wire. The online learning control framework performs an inverse learning of the plant based solely on the input and output signals, and uses this information to generate control inputs. Thus, an explicit representation of the plant is not required and as a result from online learning, the controller adapts to changes of the plant. The specific method with which the inverse learning is employed, is by the use of echo-state networks (ESN). ESNs are a class of recurrent neural networks and are distinguished by their large number of hidden nodes often referred to as a dynamic reservoir. While the originally proposed method of constructing this dynamic reservoir relies on a stochastic sampling process, recent studies have suggested that using a simple and deterministic reservoir also provides sufficient performance. Here, we also investigate the impact of using such simple and deterministic reservoir structure within the online learning control framework. Experiments of the online learning control framework conducted on an SMA wire are presented.

Adaptive Control for a VTOL UAV Operating Near a Wall

In this paper, we proposed a neural-network-based adaptive controller to adapt against unknown, possibly time varying, external disturbances for a quadrotor UAV. The proposed algorithm can estimate uncertain forces, which occurs when ying in narrow areas, near walls and/or other surfaces, so that the controller can maintain satisfactory position tracking performance despite these disturbances. A proof for stability with the proposed algorithm is provided. Experimental results show that, with the proposed NN based adaptation algorithm, position tracking performance of quadrotor UAV shows satisfactory improvement in presence of external disturbances.

H-infinity control via scenario optimization for handling and stabilizing vehicle using AFS control

Due to the day by day emergence of automobile industry, improvement in sensor technologies and the need to reduce the occurrence of accident vehicle handling and stability improvement technologies emanating. In this paper a novel H infinity controller via scenario optimization (H-infinity VSO) is presented for Active front steer (AFS) control the aim is to handle and stabilize vehicle under the uncertainty of parameter road adhesion coefficient µ. Parameter uncertainty is translated into perturbed matrixes. Optimal controller gains are calculated by considering predefined scenarios extracted according to the parameters ε and β. ε represents the probability of violation of LMI constraints and β represents confidence or risk failure. In this manner a chance constrained problem is solved which results in less conservativeness in stability and optimality. The results of proposed approach are much improved as compared to the optimal guaranteed cost controller (OGCC) and optimal coordination (OC) controller.

Control of a ball-bot using a PSO trained neural network

A ball-bot is an extremely agile mobile robotic platform due to its inherent instability. In order to maneuver at high speeds, a specialized controller is needed. A ball-bot can be modelled as two decoupled, 2-DOF pendulum on a cart systems. These systems comprise a classical and frequently encountered problem in the area of control theory. This paper proposed a novel technique for adaptive control of a ball-bot based on inverted pendulum on a cart system using particle swarm optimization (PSO) trained neural network. The generic PID controller is used to control the above mentioned system. The controller is able to learn the demonstrative behavior and keep the pendulum up right when subjected to perturbations. Mean Square Error for training data is found to be 7.68×10-3 and 5.5×10-4 for the testing data. The results show a promising future of the proposed technique.

On a notion of estimation entropy for stochastic hybrid systems

The advent of emerging fields such as networked control systems and wireless sensor networks involving information flow over bandwidth-constrained digital channels has presented the control and information theory communities with a new set of challenges. One of these challenges is to accurately estimate the trajectory of (stochastic) nonlinear systems using sampled and quantized state measurements transmitted over a finite-bandwidth digital channel. This has led researchers to study these problems by combining notions from both control and communication domains, two fields which have been traditionally treated separately. A recent notion introduced in this context is called estimation entropy, which has been defined as the minimum bit rate required to estimate the trajectory of a deterministic nonlinear system with a given exponential convergence rate. In this paper, we show that this notion can be extended to continuous-time stochastic systems as well, particularly, a class of stochastic hybrid systems. We also provide an upper bound on the estimation entropy for this class of systems.

Scenario based robust control design for uninterruptible power supplies

The paper presents the designing of a robust control scheme for uninterruptible power supplies (UPS) by implementing scenario based approach which is a probabilistic solution framework to synthesize and analyze the problems that can be expressed in the form of minimization of a linear objective subject to convex constraints parameterized by uncertainty terms. The scenario based problem being a standard convex optimization problem possesses the solution which is approximately feasible for the infinite set of constraints and is obtained by appropriate sampling of uncertain constraints. The load variation of UPS affects the output voltage and hence is considered as a linear uncertain parameter with a known range in the dynamic model of UPS. An LPV H-infinity optimization problem is formulated in the presence of load variations maintaining the output voltage to the desired value. To design the control problem guaranteeing an a-priori particular probabilistic robustness we use the scenario based approach by randomly selecting the finite number of uncertain scenarios for the load variations instead of using conventional ploytopic approach which covers and comments on the entire uncertainty set for UPS and the solution of scenario problem is computed efficiently by solving the optimization problem using linear matrix inequalities (LMIs). Numerical simulations and comparisons with LPV H-infinity and with linear PID control design shows the proposed scheme makes the system more robust towards uncertainties as well as exogenous disturbances to make it stable and improving its THD and transient performance.

Scenario based approach for control design for DC-DC Buck Converter

The article presents the application of a probabilistic robust control scheme for DC-DC Buck Converter, on the basis of the idea of randomly selected scenarios from an uncertain set. This technique involves the analysis of a wide range of control system analysis and design problems for uncertain systems which are amenable to solve the numerical problems in efficient way if the requirements regarding robustness are imposed in probabilistic sense within the paradigm. The output voltage of the converter can be affected by changes in load current. This variation in load appears as a linear uncertain parameter, with a known range, in the dynamic model of the converter. Maintaining the output voltage to the desired value in the presence of load variations is formulated as an LPV H-infinity optimization problem. However, instead of covering the entire uncertainty set using the conventional polytopic approach, we use the scenario based approach to extract the scenarios by explicitly and providing the bound on the number of scenarios, resulting in a design to guarantee an a-priori particular probabilistic robustness in design of control problem for DC-DC Buck Converter. This results in a marginal improvement in transient performance, with regard to overshoot and settling time, of the closed loop system. Numerical simulations are performed to demonstrate the above, and a comparison with the LPV polytopic approach and linear PID control is presented.

Variable robustness control design for system of aerospace vehicle

A robust control is developed for the system of aerospace vehicle via variable robustness control (VRC) algorithm. Aerospace vehicle dynamics contain uncertainties due to cross coupling between longitudinal and lateral dynamics. The proposed scheme makes the system of aerospace vehicle impervious of uncertainties ensuring high performance. The variable robustness control technique is based upon scenario approach of robust control design which gives flexibility of modulating robustness in control design. Scenario approach is a probabilistic approach in which cost function can be guaranteed with a certain probability. The effectiveness of proposed control design is established in presence of unknown non-linear dynamics. VRC uses convex optimization of constraints formulated in the form of linear matrix inequalities. The performance of proposed technique is tested in comparison with H∞ control design technique.

Adaptive control of air-to-air missile using multilayer NN feedforward and RISE feedback terms

In this work, we present the application of a neural network (NN) based adaptive control scheme for the purpose of trajectory tracking of an air-to-air missile. The nonlinear dynamic model of such missiles contain numerous aerodynamic coefficients. It is well known that uncertainty in these coefficients can cause degradation of control performance. In this work, we explore the feasibility of applying a nonlinear adaptive controller employing NN as feedforward, augmented with a continuous robust integral of the signum of the error (RISE) term connected as feedback in order to improve the performance of the degraded system. Typically NN based adaptive controller guarantees only uniformly ultimately bounded stability while the suggested control system assures semi-global asymptotic tracking of the missile. In order to accomplish the adaptive behavior, the neural network is passed through learning stage to update the weight matrices and gains by using the adaptation rules that are derived from the Lyapunov stability techniques. Results of the simulations of the proposed algorithm on six degree of freedom nonlinear missile validates feasibility of the control law. The result of the proposed controller are compared with results obtained using the backstepping technique in this paper.

Design of a novel Simulink model for surface electromyographic (SEMG) sensor design for prosthesis control

This paper presents design, simulation and fabrication of a surface electromyographic (SEMG) sensor for control of prosthetic devices. EMG activity is mainly the generation of a bio-potential signal (electrical signals) due to muscle action. These signals picked from motor points of muscles are contaminated with various intrinsic/extrinsic noises, which must be removed through different filtering techniques in order to develop a sensor that has a high signal-to-noise ratio. Power spectral density (PSD) of any EMG signal plays a vital role to determine the signal strength. A novel Simulink model has been developed which mimics various elements of an active SEMG sensor. By using this model, low/high/notch filters are designed and optimized. The model is also used to simulate the effects of these filters on power spectral density (PSD) of the EMG signal. Simulation of double rectification and smoothing (envelopment) is also carried out. EMG signals recorded from the Tibialis anterior, monopolar needle, and fine wire isometric contraction were used in this simulation. Finally, on the basis of simulation results, instrumentation of surface EMG sensor is designed and fabricated. Performance/results of developed SEMG sensor are in accordance with the simulation results of the developed Simulink model.