A neuroadaptive control scheme for elastic-joint robots is proposed that uses a relatively small neural network. Stability is achieved through standard Lyapunov techniques. For added performance, robust modifications are made to both the control law and the weight update law to compensate for only approximate learning of the dynamics. The estimate of the modeling error used in the robust terms is taken directly from the error of the network in modeling the dynamics at the currant state. The neural network used is the CMAC-RBF Associative Memory (CRAM), which is a modification of Albus's CMAC network and can be used for robots with elastic degrees of freedom. This results in a scheme that is computationally practical and results in good performance.

The paper discusses some practical problems of contact dynamics. Modelling the dynamics of contact tasks is carried out in a completely general way. Two dynamic systems, active robot system and passive environment system are brought into contact and the relevant dynamics are analyzed. The effects are: rigid-body contact force, elastodynamics in contact zone, friction in contact points, etc. Simultaneous stabilization of contact force and position is obtained using New Dynamic Position/Force Control. The general model is then applied to some more concrete problems and the simulation results are presented.

Autonomous walking bipedal machines, possibly useful for rehabilitation and entertainment purposes, need a high energy efficiency, offered by the concept of ‘Passive Dynamic Walking' (exploitation of the natural dynamics of the robot). 2D passive dynamic bipeds have been shown to be inherently stable, but in the third dimension two problematic degrees of freedom are introduced: yaw and roll.

We propose a design for a 3D biped with a pelvic body as a passive dynamic compensator, which will compensate for the undesired yaw and roll motion, and allow the rest of the robot to move as if it were a 2D machine. To test our design, we perform numerical simulations on a multibody model of the robot. With limit cycle analysis we calculate the stability of the robot when walking at its natural speed.

The simulation shows that the compensator, indeed, effectively compensates for both the yaw and the roll motion, and that the walker is stable.

An Automated Digging Control System (ADCS) for a wheel loader is developed that utilizes a behavior-based control structure combined with fuzzy logic. This controller exhibits the real-time reactive responses necessary for executing digging tasks in an uncertain, unstructured and dynamic excavation environment. This paper presents field test results of a prototype ADCS that was developed and implemented on a Caterpillar 980G wheel loader. Test results show that the performance of the automated system is comparable to that of an expert human operator in a wide range of excavation situations.

This paper presents a real-time algorithm for modifying the trajectory of a manipulator approaching a moving target. The algorithm is based on the superposition scheme; a model developed based on human motion behavior. The algorithm generates a smooth trajectory toward the new target by calculating the vectorial sum between the first trajectory (initial position and first target) and second trajectory (between first and second target location). The algorithm searches for the switch hme that will result in a minimum time trajectory. The idea of the algorithm is to define some domain where the optimal switching time can be found, reduce this domain as much as possible to decrease the number of the points that must be checked and try every remaining candidate in this domain to find numerically the best (optimal) switch time. The algorithm was implemented on an Adept-one robotic system taking into account velocity constraints. The actual velocity profile was found to be less smooth than specified by the mathematical model. When the switch occurs at the middle of the trajectory when the speed is close to its maximum, the change in the movement direction is performed more gently.

Robot manipulators, which are nonlinear structures and have uncertain system parameters, are complex dynamically when operated in an unknown environment. To compensate for estimate errors of the uncertain system parameters and to accomplish the desired trajectory tracking, nonlinear robust controllers are appropriate. However, when estimation errors or tracking errors are large, they require large input torques, which may not be satisfied due to torque limits of actuators such as driving motors. As a result, their stability cannot be guaranteed. In this paper, a new robust control scheme is presented to solve stability problems and to achieve fast trajectory tracking of uncertain robot manipulators in the presence of torque limits. By using fuzzy logic, new desired trajectories which can be reduced are generated based on the initial desired trajectory, and torques of the robust controller are regulated so as to not exceed torque limits. Numerical examples are shown to validate the proposed controller using an uncertain two degree-of-freedom underwater robot manipulator.

This paper presents some recent applications of sensing, guidance and telerobotic technology in the coal mining industry. Of special interest is the development of semi or fully autonomous systems to provide remote guidance and communications for coal mining equipment. We consider the use of radar and inertial based sensors in an attempt to solve the horizontal and lateral guidance problems associated with mining equipment automation. We also describe a novel teleoperated robot vehicle with unique communications capabilities, called the Numbat, which is used in underground mine safety and reconnaissance missions.

This paper presents the work on developing a sliding-leg tripod as a programmable add-on device for manufacturing. The purpose is to enhance the capabilities of any machine by providing it with a more flexible range of motion. This device can be used as a toolhead for CNC machine tools and robots, or as a work stage for coordinate measuring machines and laser scanning systems. In this paper, system modelling, analysis and control of this device is presented. System modeling includes mobility study, kinematic model and inverse kinematics. System analysis includes workspace analysis, transmission ratio and stiffness analysis. System control includes path planning, joint space control and Cartesian space prediction. It is shown that the proposed device can provide flexibility and dexterity to machines.

The design and implementation of adaptive control for nonlinear unknown systems is extremely difficult. The nonlinear adaptive control for assembly robots performing a peg-in-hole insertion is one such an example. The recently intensively studied neural networks brings a new stage in the development of adaptive control, particularly for unknown nonlinear systems. The aim of this paper is to propose a new approach of hybrid force position control of an assembly robot based on artificial neural networks systems. An appropriate neural network is used to model the plant and is updated online. An artificial neural network controller is then directly evaluated using the updated neuro model. Two control structures are proposed and the stability analysis of the closed-loop system is investigated using the Lyapunov method. Experimental results demonstrate that the identification and control schemes suggested in this paper are efficient in practice.

It has been shown that compass-like bipeds can walk passively by gravity-induced motion. Similarly, it has been stated that the unipodal phase of human gait may be regarded as a passive motion. This paper is aimed at determining precisely to what extent the swing phase of human-like walking can be considered as a non-actuated motion. More specifically, a method is developed that makes it possible to adjust initial conditions, step length and walk speed, as well as mass and inertia distribution of a bipedal robot, in order to generate a passive swing motion. The approach is demonstrated by using a 7-link planar biped. Numerical simulations show that purely passive unipodal transferring cannot be fully satisfactory. This approach is further considered in Part 2 of the paper which consists in computing optimal actuating joint torques to create perfectly feasible motions.

This paper presents the kinematic and dynamic modeling of a two degrees of freedom manipulator attached to a vehicle with a two degrees of freedom suspension system. The vehicle is considered to move with a constant linear speed over an irregular ground-surface while the end-effector tracks a desired trajectory in a fixed reference frame. In addition, the effects of highly coupled dynamic interaction between the manipulator and vehicle (including the suspension system's effects) have been studied. Finally, simulation results for the end-effector's straight-line trajectory are presented to illustrate these effects.

We present a data fusion control scheme for the hand-held camera of the SCORBOT-ER VII robot arm for learning visual tracking and interception. The control scheme consists of two modules: The first one generates candidate actions to drive the end-effector as accurate as possible directly above a moving target, so that the second module can handily take over to intercept it. The desired camera-joint coordinate mappings are generalized by Elman neural networks for a tracking module. The intercept module then determines a suitable intercept trajectory for the robot within the required conditions. The simulation results support the claim that it could be successfully applied to track and intercept a moving target.

This paper presents an adaptive steering controller for achieving accurate and prompt steering control with noisy steering command signals and drifting valve characteristics on an automated agricultural tractor with an electrohydraulic steering system. It is difficult to accomplish performance objectives with conventional PID controllers because of the effects of disturbances and unknown factors. The adaptive controller, consisted of an adaptive gain tuner and an adaptive nonlinearity compensator, was to overcome these performance obstacles. Test results indicated that this controller provided an effective means for achieving satisfactory steering control for automated tractor traveling on changing and unpredictable farm field courses.

The paper addresses a problem of “in-hand” locating parts of different shapes in robotic grasping. The goal of the process is to locate a part of an arbitrary shape from an imprecisely determined initial position within a gripper to a final prescribed one. Two possible approaches to solve the problem are considered: non-adaptive, using ordinary rigid jaws of gripper and, adaptive, using an adaptive jaw which improves the performance of the locating process. The latter approach is proposed to be solved by a new type of grasping mechanism. Its theoretical analysis enables to obtain formal conditions for part behavior during the successive steps of the locating process. This process was simulated and then experimentally investigated on an actual gripper model. The proposed new class of mechanisms opens a promising avenue to the creation of a practical class of universal robotic grippers for industry.

In this paper a new approach for the formulation of the friction forces velocity function is introduced. The scope of this formulation is to facilitate the implementation of control laws for systems where friction forces appear. The friction model includes the exponential decay part, the Coulomb and viscous friction. The introduced formulation is based on the observation that the friction coefficient function of velocity can be presented as the solution of a linear differential equation. Due to this linearity, the parameters of the derived differential equation can be estimated easily by an adaptive system. The estimation of these parameters is equivalent to the estimation of the friction coefficient in the full range of operational velocities. This knowledge gives to the designed control systems the potential to avoid successfully the stick-slip phenomenon.

A control law for one D.O.F. system, where friction appears, is designed in order to prove the applicability of the proposed formulation of the friction model in control systems. A MRAC adaptive algorithm estimates the differential friction model parameters, using the measured friction force, while a sliding controller adjusts the motion of the mechanical system. The proposed friction model can be used in any control system where friction forces have to be compensated. The linear form of the model is suitable for common adaptive estimators. Therefore, the proposed structure is suitable for robotic applications, such as assembly, deburring, etc.

In this work another perturbation estimation sliding mode based control algorithm is introduced for a class of robotic systems in the presence of structured and unstructured uncertainties and external disturbances. The effects of these uncertainties are combined into a single quantity. A full order device with the actuator voltages as control inputs is assumed in control design. The decentralized control scheme with only a partial state feedback is applied. A modification of the switching functions with perturbation estimation is introduced. The salient features of this approach is that the perturbations are effectively treated by a computationally straightforward procedure. The proposed controller is applied to a minimal configuration direct drive robot mechanism.

In this paper a method for the automatic generation of dynamics for modular robots is presented. A modular reconfigurable robot consists of link modules and joint modules of various specifications. We analyze the abstract architecture of modular robots. A total of nine types of connecting forms and three types of joint forms are identified with reference to their own systems. The geometric relationships are derived by the group theory. According to the modular idea, the formulations of the velocity, acceleration and other essential equations of the link module and the joint module are formed compensatively and recursively. Then the dynamic model is generated automatically by a compensative and recursive method.

A class of three-legged modular reconfigurable parallel robots is designed and constructed for precision assembly and light machining tasks by using standard active and passive joint modules in conjunction with custom designed links and mobile platforms. Since kinematic errors, especially the assembly errors, are likely to be introduced, kinematic calibration becomes particularly important to enhance the positioning accuracy of a modular reconfigurable robot. Based on the local frame representation of the Product-Of-Exponentials (Local POE) formula, a self-calibration method is proposed for these three-legged modular reconfigurable parallel robots. In this method, both revolute and prismatic joint axes can be uniformly expressed in twist coordinates by their respective local (body) frames. Since these local frames can be arbitrarily defined on their corresponding links, we are able to calibrate them, and yet retain the nominal local description of their respective joints, i.e., the nominal twist coordinates and nominal joint displacements, to reflect the actual kinematics of the robot. The kinematic calibration thus becomes a procedure of fine-tuning the locations and orientations of the local frames. Using mathematical tools from differential geometry and group theory, an explicit linear calibration model is formulated based on the leg-end distance errors. An iterative least-square algorithm is employed to identify the error parameters. A simulation example of calibrating a three-legged (RRRS) modular parallel robot shows that the robot kinematics can be fully calibrated within two to three iterations.

We present a feature-based probabilistic map building algorithm which directly utilizes time and amplitude information of sonar in indoor environments. Utilizing additional amplitude-of-signal (AOS) obtained concurrently with time-of-flight (TOF), the amount of inclination of target can be directly calculated from a single echo, and the number of measurements can be greatly reduced with result similar to dense scanning. A set of target groups (set of hypothesized targets originated from one measurement) is used and refined by each measurement using an extended Kalman filter and Bayesian conditional probability. Experimental results in a real indoor environment are presented to show the validity of our algorithm.

Only one parallel wrist with three equal legs containing just revolute pairs has been already presented in the literature. This parallel wrist is overconstrained, i.e., it involves three degrees of freedom required to orientate the end effector by using repetitions of constraints. The overconstrained mechanisms have the drawback of jamming or undergoing high internal loads when geometric errors occur. This paper presents a new parallel wrist, named 3-RUU wrist. The 3-RUU wrist is not overconstrained. It has three equal legs just involving revolute pairs and actuators adjacent to the frame and uses an architecture (3-RUU) already employed to obtain manipulators that make the end effector translate. The 3-RUU wrist kinematic analysis is addressed. This analysis shows that the new parallel wrist can reach singular configurations (translation singularities) in which the spherical constraint between end effector and frame fails. The singularity condition that makes finding all the 3-RUU wrist singular configurations possible is written in explicit form and geometrically interpreted.