The CyberWalk is a large size 2D omni-directional platform that allows unconstrained locomotion possibilities to a walking user for VR exploration. In this paper we present the motion control design for the platform, which has been developed within the homonymous European research project. The objective is to compensate the intentional motion of the user, so as to keep her/him always close to the platform center while limiting the perceptual effects due to actuation commands. The controller acts at the acceleration level, using suitable observers to estimate the unmeasurable intentional walker's velocity and acceleration. A moving reference position is used to limit the accelerations felt by the user in critical transients, e.g., when the walker suddenly stops motion. Experimental results are reported that show the benefit of designing separate control gains in the two orthogonal directions (lateral and sagittal) of a frame attached to the walker.

The CyberCarpet is an actuated platform that allows unconstrained locomotion of a walking user for VR exploration. The platform has two actuating devices (linear and angular) and the motion control problem is dual to that of nonholonomic wheeled mobile robots. The main control objective is to keep the walker close to the platform center. We first recall global kinematic control schemes developed at the velocity level, i.e., with the linear and angular velocities of the platform as input commands. Then, we use backstepping techniques and the theory of cascaded systems to move the design to control laws at the acceleration level. Acceleration control is more suitable to take into account the limitations imposed to the platform motion by the actuation system and/or the physiological bounds on the human walker. In particular, the availability of platform accelerations allows the analytical computation of the apparent accelerations felt by the user.

Exploration of virtual worlds with unconstrained locomotion possibilities for the user is the main objective of the European research project CyberWalk. This should be achieved through the use of an actuated platform (the CyberCarpet) that compensates for the walker's locomotion in such a way to keep her/him close to the platform center. This paper presents the control problem for the platform motion, including objectives and constraints, overall control architecture, and kinematic modeling. Since the platform has only two actuating devices (linear and angular), the control problem is similar to that of output regulation for nonholonomic wheeled mobile robots in the presence of an unpredictable disturbance due to walker's locomotion. Based on the kinematic model, a velocity control design achieving input-output decoupling and linearization is proposed and its performance is verified by simulations

We consider the problem of real-time detection of collisions between a robot manipulator and obstacles of unknown geometry and location in the environment without the use of extra sensors. The idea is to handle a collision at a generic point along the robot as a fault of its actuating system. A previously developed dynamic FDI (fault detection and isolation) technique is used, which does not require acceleration or force measurements. The actual robot link that has collided can also be identified. Once contact has been detected, it is possible to switch to a suitably defined hybrid force/motion controller that enables to keep the contact, while sliding on the obstacle, and to regulate the interaction force. Simulation results are shown for a two-link planar robot.

We present a scheme for identifying the time profile of actuator faults that may affect a robot manipulator. Starting from our previous method for fault detection and isolation (FDI) based on generalized momenta, fault identification is additionally obtained through the H/sub /spl infin//-design of a state observer for uncertain systems. For each separate fault channel, the identifier consists of a linear filter driven by the corresponding residual signal. Under the weak assumption of bounded time derivative for the otherwise unknown fault input to be identified, the fault estimation error is shown to be ultimately uniformly bounded, with ultimate bound that can be set arbitrarily small. The information on the type and severity of the fault may then be used for reconfiguring the control strategy. Experimental results on a 2R planar manipulator are presented.

An adaptive scheme is presented for actuator fault detection and isolation (FDI) in robotic systems, based on the use of generalized momenta and of a suitable overparametrization of the uncertain robot dynamics. This allows to obtain an accurate and reliable detection and isolation of possibly concurrent faults also during the parameter adaptation phase. Experimental results are reported for a planar robot under gravity, considering partial, total, or bias-type failures of the motor torques.

We present a method based on the use of generalized momenta for detecting and isolating actuator faults in robot manipulators. The FDI scheme does not need acceleration estimates or simulation of the nominal robot dynamics and covers a general class of input faults. Numerical results for a 2R robot undergoing also concurrent actuator faults are reported. This method is extended to robots with joint elasticity and to the inclusion of actuator dynamics.

On-line clustering is required whenever huge amounts or continuous flows of data must be classified and/or described in a compact way. Most available unsupervised competitive learning methods exhibit the problem that the units of the self-organizing map "accumulate" in the regions of the sample space characterized by higher data density, while in clustering it is generally desired that just one node represents each data cluster, independently of the relative density of points in the different clusters. This paper presents a novel on-line, hard-competitive algorithm that deals with this basic limitation, showing very good performance in clustering data from artificial distributions, as well as real data within the problem of motion-based scene segmentation in automated video surveillance.

In many industrial applications, robotic systems accomplish the task of sorting items on moving conveyor belts. The list of objects to be gripped can be viewed as a queue of clients waiting to be served. The main peculiarities of this queue are that the serving times of its elements vary in a dynamic way, and that any client has to be served before it exits the robot workspace. In most practical cases, a simple first-in-first-out (FIFO) rule can be used for scheduling the jobs in the queue without dealing at all with the above issues. However, there are situations of industrial interest, as in the automatic sorting of wasted material, where the stochastic behavior of items flow gives rise to repeated overload situations, where the FIFO rule performs very inefficiently, requiring different scheduling strategies. In this paper, we propose two innovative online scheduling rules, based on suitable modifications of standard strategies for static queues, having the same complexity, but improved performance in the considered dynamic case. Simulation results confirm the validity of the proposed techniques.

A particular class of underactuated systems is obtained by considering kinematically redundant manipulators for which all joints are passive and the only available inputs are forces/torques acting on the end-effector. Under the assumption that the degree of redundancy is provided by prismatic joints located at the base, we address the problem of steering the robot between two arbitrary equilibrium configurations. By performing a preliminary partial feedback linearization, the dynamic equations take a convenient triangular form, which is further simplified under additional hypotheses. We give sufficient conditions for controllability of this kind of mechanisms. With a PPR robot as a case study, an algorithm is proposed for computing end-effector commands that produce the desired reconfiguration in finite time. Simulation results and a discussion on possible generalizations are given.

We outline a general approach for the stabilization of robots with passive joints, an interesting example of mechanical systems that may not be controllable in the first approximation. The proposed method is based on a recently introduced iterative steering paradigm, which prescribes the repeated application of a contracting open-loop control law. In order to complete efficiently such a law, the dynamic equations of the robot are put in a suitable form, via partial feedback linearization and approximate nilpotentization. The design procedure is illustrated for a 2R robot moving in the horizontal plane with a single actuator at the base. Experimental results are presented for a laboratory prototype.

The authors analyze the dynamic mobility of a kinematically redundant robot driven by forces/torques imposed on the end-effector, an interesting example of underactuated system. Under suitable assumptions, the system can be put via feedback in two special forms, namely the second-order triangular and Caplygin forms. Nonlinear controllability tools are used to derive conditions under which the robot can be steered between two given configurations using end-effector commands. With a PPR robot as a case study, a steering algorithm is proposed that achieves reconfiguration in finite time.

A general control-oriented formalism has been introduced by DeLuca and Manes (1991, 1994) to describe robot-environment interaction in the case of a single robot in contact with a possibly dynamic environment. Two possibilities were obtained in the design of hybrid motion-force control laws depending on whether motion or force is explicitly controlled along properly defined dynamic directions. An extension of this formalism for modeling and controlling cooperating robots in the manipulation of a payload is presented. Several contact arrangements and environmental constraints on the object can be considered, mixing the presence of kinematic constraints and dynamic interactions. The modeling approach leads to the characterization of complementary directions in the task space: those where only end-effector velocities are admissible, those where only reaction forces may exist, and those in which energy can be transferred between each robot and the payload. Two classes of model-based hybrid force-motion controllers are then designed similarly to the single robot case. The authors show that the classical approach of controlling payload motion and internal forces for cooperating robots is recovered as one special case. Moreover, within this framework a new control alternative naturally arises, namely the possibility of controlling both internal forces and active forces producing motion. Numerical simulation results are reported for power grasp and hard finger point contacts and with the two alternative control laws.