Seul Jung

Hardware Implementation of a Real-Time Neural Network Controller With a DSP and an FPGA for Nonlinear Systems

In this paper, we implement the intelligent neural network controller hardware with a field programmable gate array (FPGA)-based general purpose chip and a digital signal processing (DSP) board to solve nonlinear system control problems. The designed intelligent control hardware can perform real-time control of the backpropagation learning algorithm of a neural network. The basic proportional-integral-derivative (PID) control algorithms are implemented in an FPGA chip and a neural network controller is implemented in a DSP board. By using a high capacity of an FPGA chip, the additional hardware such as an encoder counter and a pulsewidth modulation (PWM) generator is implemented in a single FPGA chip. As a result, the controller becomes cost effective. It was tested for controlling nonlinear systems such as a robot finger and an inverted pendulum on a moving cart to show performance of the controller

Force tracking impedance control of robot manipulators under unknown environment

In this paper, a new simple stable force tracking impedance control scheme that has the capability to track a specified desired force and to compensate for uncertainties in environment location and stiffness as well as in robot dynamic model is proposed. The uncertainties in robot dynamics are compensated by the robust position control algorithm. After contact, in force controllable direction the new impedance function is realized based on a desired force, environment stiffness and a position error. The new impedance function is simple and stable. The force error is minimized by using an adaptive technique. Stability and convergence of the adaptive technique are analyzed for a stable force tracking execution. Simulation studies with a three link rotary robot manipulator are shown to demonstrate the robustness of the proposed scheme under uncertainties in robot dynamics, and little knowledges of environment position and environment stiffness. Experimental results are carried out to confirm the proposed controllers performance.

Control Experiment of a Wheel-Driven Mobile Inverted Pendulum Using Neural Network

The mobile inverted pendulum is developed and tested for an intelligent control experiment of control engineers. Intelligent control algorithms are tested for the control experiment of a low cost mobile inverted pendulum system. Online learning and control using neural network of a wheel-driven mobile inverted pendulum system is presented. Neural network learning algorithm is embedded on a digital signal processing board along with primary proportional-integral-differential controllers to achieve real time control. Without knowing dynamics of the system, uncertainties in system dynamics are compensated by neural network in an online fashion. Digital filters are designed for a gyro sensor to compensate for a phase lag. Experimental studies of balancing the pendulum and tracking the desired trajectory of the cart for one dimensional motion are conducted. Results show the robustness of the proposed controller even when outer impacts as disturbance are present.

Neural network impedance force control of robot manipulator

The performance of an impedance controller for robot force tracking is affected by the uncertainties in both the robot dynamic model and environment stiffness. The purpose of this paper is to improve the controller robustness by applying the neural network (NN) technique to compensate for the uncertainties in the robot model. NN control techniques are applied to two impedance control methods: torque-based and position-based impedance control, which are distinguished by the way of the impedance functions being implemented. A novel error signal is proposed for the NN training. In addition, a trajectory modification algorithm is developed to determine the reference trajectory when the environment stiffness is unknown. The robustness analysis of this algorithm to force sensor noise and inaccurate environment position measurement is also presented. The performances of the two NN impedance control schemes are compared by computer simulations. Simulation results based on a three-degrees-of-freedom robot show that highly robust position/force tracking can be achieved in the presence of large uncertainties and force sensor noise.

Robust neural force control scheme under uncertainties in robot dynamics and unknown environment

The original impedance function is known to lack robustness due to unknown robot dynamic model and the environment. In order to improve that result, a new impedance function is derived which specifies a desired force directly. This results in a new robust robot force tracking impedance control scheme, which employs a neural network as a compensator to cancel out all uncertainties. The proposed neural force control scheme is capable of making the robot track a specified desired force as well as of compensating for uncertainties in environment location and stiffness, and in robot dynamics. Separate training signals for free-space motion and contact-space motion control are developed to train the neural compensator online. The design of the training signals is justified. Simulation studies with a three-link rotary robot manipulator are carried out and the results show excellent force tracking performance.

Gyro sensor drift compensation by Kalman filter to control a mobile inverted pendulum robot system

In this paper, a sensor fusion technique of low cost sensors such as a gyro sensor and a tilt sensor to measure the balancing angle of the inverted pendulum robot system accurately is implemented. The complimentary filter consisting of the lowpass filter for the gyro sensor and the highpass filter for the tilt sensor are used based on the frequency response characteristics of those sensors. The Kalman filter is used to estimate the angle based on filtered sesnor data. Experimental studies of balancing and position control of the mobile inverted pendulum robot system are conducted to validate the proposed estimation method.

Neural Network Control for Position Tracking of a Two-Axis Inverted Pendulum System: Experimental Studies

In this paper, experimental studies of a decentralized neural network control scheme of the reference compensation technique applied to control a 2-degrees-of-freedom (2-DOF) inverted pendulum on an x-y plane are presented. Each axis is controlled by two separate neural network controllers to have a decoupled control structure. Neural network controllers are applied not only to balance the angle of pendulum, but also to control the position tracking of the cart. The decoupled control structure can compensate for uncertainties and cancel coupling effects. Especially, a circular trajectory tracking task is tested for position tracking control of the cart while maintaining the angle of the pendulum. Experimental result shows that position control of the inverted pendulum and cart is successful.

Force Tracking Impedance Control for Robot Manipulators with an Unknown Environment: Theory, Simulation, and Experiment

In impedance control for force tracking, it is well known that the reference trajectory of the robot is calculated from known environmental stiffness. The authors present a simple technique for determining the reference trajectory under the condition that the environment is unknown. The technique is developed based on the replacement of the unknown stiffness with a function of the measured force. Combining this technique with the impedance function yields the force tracking impedance function. Robot dynamic uncertainties are assumed to be compensated by a robust position control method based on time-delayed control. The local stability at equilibrium points is analyzed with respect to uncertainty in environmental position. Computer simulation studies demonstrate that force tracking using the proposed technique is excellent for unknown environment and dynamics uncertainty. The practicality of the technique is also verified experimentally using a PUMA 560 manipulator.

Experimental studies of neural network impedance force control for robot manipulators

In this paper, the neural network force control is presented. Under the framework of impedance control, neural network is used to compensate for all the uncertainties from robot dynamics and unknown environment. A modified simple impedance function is realized after the convergence of the neural network. Learning algorithms for the neural network to minimize the force error directly are designed. As a test-bed, the large X-Y table robot was implemented. Experimental results obtained show better force tracking when the neural network is used.

Neural network inverse control techniques for PD controlled robot manipulator

In this paper neural network (NN) control techniques for non-model based PD controlled robot manipulators are proposed. The main difference between the proposed technique and the existing feedback error learning (FEL) technique is that compensation of robot dynamics uncertainties is done outside the control loop by modifying the desired input trajectory. By using different NN training signals, two NN control schemes are developed. One is comparable to that in the FEL technique and another has to deal with the Jacobian of the PD controlled robot dynamic system. Performances of both controllers for various trajectories with different PD controller gains are examined and compared with that of the FEL controller. It is shown that the new control technique performed better and robust to PD controller gain variations.