Neural Network Direct Control with Online Learning for Shape Memory Alloy Manipulators

Abstract

New actuators and materials are constantly incorporated into industrial processes, and additional challenges are posed by their complex behavior. Nonlinear hysteresis is commonly found in shape memory alloys, and the inclusion of a suitable hysteresis model in the control system allows the controller to achieve a better performance, although a major drawback is that each system responds in a unique way. In this work, a neural network direct control, with online learning, is developed for position control of shape memory alloy manipulators. Neural network weight coefficients are updated online by using the actuator position data while the controller is applied to the system, without previous training of the neural network weights, nor the inclusion of a hysteresis model. A real-time, low computational cost control system was implemented; experimental evaluation was performed on a 1-DOF manipulator system actuated by a shape memory alloy wire. Test results verified the effectiveness of the proposed control scheme to control the system angular position, compensating for the hysteretic behavior of the shape memory alloy actuator. Using a learning algorithm with a sine wave as reference signal, a maximum static error of 0.83° was achieved when validated against several set-points within the possible range.

Review of Neural Network Modeling of Shape Memory Alloys

“…However, the research does not end at this point; the maximization of the performance of the chosen NN is the next step. Different methods can be used including the following AI algorithms: Marquardt [ 80 ], Jordan–Plus–Elman NARX [ 86 ], back-propagation [ 10 , 81 , 91 , 98 , 99 ], batch back-propagation (BPP) [ 98 ], Quick Prop QP, incremental back-propagation (IBP) [ 98 ], fuzzy algorithm, bacteria foraging algorithm (MBFA) [ 90 ], GRNN (general regression) [ 83 ], particle swarm optimization (PSO) [ 89 ], VIKOR fuzzy algorithm, fuzzy algorithm genetic algorithm (GA), and gradient descent algorithm (GDA) [ 90 ]. We have seen in the descriptions that using metaheuristic optimization can help enhance the results.…”

“…The first one is a variable structure control (VSC) switch controller (a switch that actuates magnetic contactors and remote-operated controllers) with a hidden layer around the switching surface of the manipulator, and the second one is a neural plant model. Moreover, in [ 10 ], backpropagation NN (BPNN) direct control was developed with online learning control. This means that actuator position data were used to update the weight coefficients of the neural network.…”

“…Since the 1980s, a new and dynamically developing field has been emerging known as intelligent materials or smart materials (SMs). These materials play an essential role in various technological fields [ 1 , 2 ], such as aerospace and automobile [ 3 ], robotics [ 4 ], medical [ 5 ], engineering [ 6 ], actuators [ 7 ], sensors [ 8 , 9 ], and rotary manipulators [ 10 ]. Shape memory alloys (SMAs) are one of the most highlighted smart materials [ 11 ].…”

Modeling the butterfly behavior of SMA actuators using neural networks

“…By these categories, NNs can provide interesting solutions to modeling and predicting the characteristics of smart materials. Concerning SMA, NN was used in various applications to model the hysteresis behavior of SMA [12], to train NNs' parameters using particle swarm optimization technique (PSO) [19], and to predict the reduction factor of SMA as a function of the reinforcement ratio and the reinforcement modulus of elasticity of SMA [20], using shallow NNs with constraints [21], and feed-forward and back-propagation NNs [22].…”

Active Modeling and Control for Shape Memory Alloy Actuators

“…Adaptive control algorithms, such as output feedback direct adaptive control [30], model reference active control [31], and robust indirect adaptive control [32], etc., were proposed to control SMA actuators. Moreover, neural network model prodictive control [33], neural network feedforward control with RISE feedback [34] and neural network direct control with online learning [35] were also used to control the modeling error of SMA actuators. However, how to reject the modeling error effectively in a wide working condition of SMA actuators is still an open problem.…”