Chow Yin Lai

Product-of-exponential (POE) model for kinematic calibration of robots with joint compliance

In this paper, a robot calibration method is proposed where the effects of joint compliance is taken into account when formulating the model. The calibration process identifies the parameters describing the relative poses of the local frames attached to the links of the robot and the parameters determining joint deflections such as joint stiffness, the mass and center of gravity of each link. A kinematic model of the robot which considers the effects of joint compliance is first formulated using the product-of-exponentials approach. Subsequently, the calibration process is described where the parameters within the robot model are updated based on measurement data. Finally, a comprehensive simulation study is conducted to validate the assumptions made during problem formulation and to verify the effectiveness of the proposed approach.

Improving the transient performance in robotics force control using nonlinear damping

Robotics force control is increasingly being used in the manufacturing industry, to perform surface polishing, deburring, parts mating etc. However, to avoid the high impact force which could damage the workpiece, a common strategy is to command the robot to approach the workpiece slowly, which increases the overall cycle time. In this paper, we propose a nonlinear damping control scheme which reduces the force overshoot without compromising on the dynamic response. The controller design and the parameter tuning are relatively straightforward because of the clear physical meanings of the parameters. Finally, the effectiveness of the proposed controller is verified through simulation studies.

Metacognitive learning approach for online tool condition monitoring

As manufacturing processes become increasingly automated, so should tool condition monitoring (TCM) as it is impractical to have human workers monitor the state of the tools continuously. Tool condition is crucial to ensure the good quality of products—worn tools affect not only the surface quality but also the dimensional accuracy, which means higher reject rate of the products. Therefore, there is an urgent need to identify tool failures before it occurs on the fly. While various versions of intelligent tool condition monitoring have been proposed, most of them suffer from a cognitive nature of traditional machine learning algorithms. They focus on the how-to-learn process without paying attention to other two crucial issues—what-to-learn, and when-to-learn. The what-to-learn and the when-to-learn provide self-regulating mechanisms to select the training samples and to determine time instants to train a model. A novel TCM approach based on a psychologically plausible concept, namely the metacognitive scaffolding theory, is proposed and built upon a recently published algorithm—recurrent classifier (rClass). The learning process consists of three phases: what-to-learn, how-to-learn, when-to-learn and makes use of a generalized recurrent network structure as a cognitive component. Experimental studies with real-world manufacturing data streams were conducted where rClass demonstrated the highest accuracy while retaining the lowest complexity over its counterparts.

General framework of the force and compliant motion control for macro mini manipulator

This paper introduces a general control framework for macro mini manipulator to improve the force and compliant motion control of robotic manipulators. RMRC (Resolved Motion Rate Control) is used as the controller for the industrial (macro) manipulator, while a switching between position control and force control is applied for the mini. The algorithm shows that it can work well in searching and contacting the workpiece, even with limited information of the location of the workpiece. It is also suitable for the compliant motion when it deals with an unknown contour of the workpiece. The framework is validated through experiments which utilize 7-DOF Mitsubishi PA-10 as the macro manipulator and 1-DOF voice coil as the mini manipulator.

Nonlinear damping for improved transient performance in robotics force control

In order to improve the productivity of processes involving contact between robotic manipulator and workpiece, it is necessary to shorten the time for the contact force to reach the desired value. In doing so, however, the contact force may experience overshoot which could damage the workpiece. In this paper, we propose a nonlinear damping control scheme which can reduce the force overshoot without compromising on the reaction speed. The controller design and the tuning of its parameters are relatively straightforward because the parameters possess clear physical meanings. Finally, experiments are carried out to verify the effectiveness of the proposed controller.

An adaptive neural network compensator for decoupling of dynamic effects of a macro-mini manipulator

In this paper, an adaptive neural network compensator is proposed to improve the control performance of a macro-mini robotic manipulator, whereby an end effector called the mini manipulator is mounted at the end of an industrial robot called the macro manipulator in robotic literature. With the macro-mini architecture, the manipulator has the advantages of a large workspace due to the industrial robot, as well as fast dynamic response and diverse functions due to the end effector with optimized mechanical design. However, the coupling dynamics between two separated systems will also influence the control performance, especially at the end effector. An adaptive neural network compensator is added to the minis original control system, and it can estimate and eliminate the dynamic coupling effect coming from the macro system in real-time in an on-line manner. Simulation shows that even if the model of the macro is unknown and its controller unchanged, in the presence of external disturbance at macro manipulator, the dynamic coupling effect can be almost eliminated in both position and force controls by the appended adaptive neural network compensator. Experimental results also demonstrate that as compared to the mini based on feedback linearization with the PID controller, vibration at the end point due to the coupling effect can be obviously suppressed using the adaptive compensator, and the steady state error is also gradually improved.

Decoupling of macro-mini manipulator using adaptive neural networks

Attaching a small manipulator (mini) with fast dynamic response at the end of a bigger manipulator (macro) with larger workspace leads to the concept of macro-mini manipulator, which is seen as a way to improve the system performance as compared to the macro manipulator acting alone, for example in terms of positioning accuracy. However, cross coupling between the two counterparts could undermine the practicality of the concept. In this paper, an adaptive neural network decoupler is presented to reduce the coupling effect of the macro-mini manipulators, without the need to have a proper dynamic model of the macro, and without alteration to the macros controller. The stability of the proposed scheme is analyzed through the use of Lyapunov criterion. Simulation results show that by using the proposed neural network decoupler, the positioning accuracy of the macro-mini system can be improved significantly even when the macro manipulator is perturbed by external disturbances.

A generalized underactuated robot system inversion method using Hamiltonian formalism

A new generalized method of stable model inversion is presented with the aim of providing solutions for the feedforward control of underactuated robots. The area of application is in SISO and MIMO systems within robotics which contain only scleronomous constraints. This generalized restriction is discussed followed by a justification of its sufficiency. The method uses a boundary value problem framework along with Hamiltonian formalism, representing the dynamic equations of motion, to solve for the stable model inversion of a robotic system. The benefits of the method include energy savings, enhanced safety, and robot simplification. An example of the robot feedforward control solution is presented to conclude the work.

Hamiltonian exploitation in underactuated robot system inversion

A new stable model inversion method extension is presented with the aim of providing solutions to the feedforward control of underactuated robots performing cyclic tasks. The method uses a boundary value problem framework along with a Hamiltonian formalism, representing the dynamic equations of motion, to find solutions for the control of movement between non-equilibrium points in taskspace. The benefits of the method are demonstrated with a simulated robot system.

Improving Positioning Accuracy Using a General Framework for Macro Mini Manipulation

This paper introduces a general framework for macro mini manipulator to improve the positioning accuracy of an industrial manipulator. RMRC (Resolved Motion Rate Control) is used as the controller for the industrial (macro) manipulator while PID with gravity compensation is used as the controller of the mini manipulator. Gain scheduling is introduced to reduce the coupling error that arises from the motions of the macro and mini manipulator. This paper also presents a trajectory planner which ensures the macro and mini manipulator will always be inside of their respective workspaces. The experiment utilizes 7-DOF Mitsubishi PA-10 as the macro manipulator and 1-DOF voice coil actuator as the mini manipulator. The experiments show that the framework improves the position accuracy.