Neural network–based sliding-mode control of a tendon sheath–actuated compliant rescue manipulator

Abstract

The novel contribution of this article is to propose a neural network–based sliding-mode control strategy for improving the position-control performance of a tendon sheath–actuated compliant rescue manipulator. Structural design of a rescue robot with slender and compliant mechanical structure is introduced. The developed robot is capable of drilling into the narrow space under debris and accommodating complicated configuration in ruins. Dynamics modeling and parameters identification of a compliant gripper with flexible tendon sheath transmission are researched and discussed. Moreover, the neural network–based sliding-mode control scheme developed based on radial basis function network is proposed to improve the position-control accuracy of the gripper with modeling uncertainties and external disturbances. The stability of the proposed control system is demonstrated using Lyapunov stability theory. Further experimental investigation including trajectory-tracking experiments and step-response experiments are conducted to confirm the effectiveness of the proposed neural network–based sliding-mode control scheme. Experimental results show that the proposed neural network–based sliding-mode control scheme is superior to cascaded proportional–integral–derivative controller and conventional sliding-mode controller in position-control application.

Design and Xilinx Virtex-field-programmable gate array for hardware in the loop of sensorless second-order sliding mode control and model reference adaptive system–sliding mode observer for direct torque control of induction motor drive

“…The OLSFE given in equation ( 25) is replaced by an SMO provided in equation ( 16). The suggested SMO integrates an adaptation mechanism given by equation (22). The stability analysis is verified using the Lyapunov function in equation (19).…”

“…[17][18][19] Moreover, the SMC has promising features, and it is utilized in several research studies for controlling the industrial systems. 20 However, the main drawback of the first-order SMC is the chattering phenomenon followed by high oscillations around the equilibrium point due to its discontinuous nature of the control action, [20][21][22] which causes vibrations in the electrical machines and within the mechanical parts of the system. This phenomenon makes this technique unsuitable for some industrial applications and leads also to its rejection by designers.…”

Control Method of Flexible Manipulator Servo System Based on a Combination of RBF Neural Network and Pole Placement Strategy

“…In the RBF neural network, the most common activation function is Gaussian [31]. The Gaussian activation function has the following advantages:…”

Sensorless direct torque control based on nonlinear integral sliding mode controllers for an induction motor drive: Experimental verification

“…Its main drawback is the presence of chattering caused by a nonlinear term present in the control law. 15,32 This phenomenon is defined by high-frequency oscillations. Several research studies have been developed to reduce this phenomenon.…”

“…33,34 In the study by Gadoue et al, 35 the authors proposed a combination of SM control and fuzzy logic in order to overcome the conventional SM control drawback. In the study by Wu et al, 32 the authors proposed a combination between neural network and SM control strategy of a tendon sheath-actuated compliant manipulator used for disaster research and rescue. A comparison has been made by the authors and demonstrates that the proposed neural network-based SM control offers lower trajectory-tracking error and better step-response characteristic relative to the conventional SM control.…”