No abstract

The last decades have seen a surge of robots working in contact with humans. However, until now these contact robots have made little use of the opportunities offered by physical interaction and lack a systematic methodology to produce versatile behaviors. Here we develop the first interactive robot controller able to understand the control strategy of the human user and react optimally to their movements. We demonstrate that combining an observer with a differential game theory controller can: induce a stable interaction between the two partners; precisely identify each other's control law; and allow them to successfully perform the task with minimum effort. Simulations and experiments with human subjects demonstrate these properties and illustrate how the new controller can induce different representative interaction strategies.

The central nervous system uses stereotypical combinations of the three wrist/forearm joint angles to point in a given (2D) direction in space. In this paper, we first confirm and analyze this Donders' law for the wrist as well as the distributions of the joint angles. We find that the quadratic surfaces fitting the experimental wrist configurations during pointing tasks are characterized by a subject-specific Koenderink shape index and by a bias due to the prono-supination angle distribution. We then introduce a simple postural model using only four parameters to explain these characteristics in a pointing task. The model specifies the redundancy of the pointing task by determining the one-dimensional task-equivalent manifold (TEM), parameterized via wrist torsion. For every pointing direction, the torsion is obtained by the concurrent minimization of an extrinsic cost, which guarantees minimal angle rotations (similar to Listing's law for eye movements) and of an intrinsic cost, which penalizes wrist configurations away from comfortable postures. This allows simulating the sequence of wrist orientations to point at eight peripheral targets, from a central one, passing through intermediate points. The simulation first shows that in contrast to eye movements, which can be predicted by only considering the extrinsic cost (i.e., Listing's law), both costs are necessary to account for the wrist/forearm experimental data. Second, fitting the synthetic Donders' law from the simulated task with a quadratic surface yields similar fitting errors compared to experimental data.

This paper presents recent results on the development and control of a microgripper based on flexure joints, fabricated by LIGA and instrumented with semiconductor strain-gauge force sensors. The microgripper is the end-effector of a workstation developed to grasp and manipulate tiny objects such as the components of a typical biomedical microdevice. The development of the force control in the microgripper is of fundamental importance in order to achieve the dexterity and sensing capabilities required to perform assembly tasks for biomedical microdevices. As a step towards the definition of the force control strategy, system identification techniques have been used to model the microgripper. Results indicate that a proportional integral (PI) controller could be used to assure, at the same time, closed-loop stability of the system, and a bandwidth suitable for the intended applications. The force control is based on strain-gauge sensors which have been integrated in the microgripper and experimentally characterized. Sensor response in the idling condition and during grasp showed that they can provide useful information for force control of the microgripper.

In this paper, an efficient charge recovery method for driving piezoelectric actuators with low frequency square waves in low power applications such as mobile microrobots is investigated. Efficiency issues related to periodic mechanical work of the actuators and the relationship among the driving electronics efficiency, the piezoelectric coupling factor, and the actuator energy transmission coefficient are discussed. The proposed charge recovery method exploiting the energy transfer between an inductor and a general capacitive load is compared with existing techniques which lead to inherent inefficiencies. A charge recovery method is then applied to piezoelectric actuators, especially to bimorph ones.Unitary efficiency can be theoretically obtained for purely capacitive loads while intrinsic losses such as hysteresis necessarily lower the efficiency. In order to show the validity of the method, a prototype driving electronics consisting of an extended H-bridge is constructed and tested by experiments and simulations. Preliminary results show that 75% of charge (i.e. more than 56% of energy) can be recovered for bending actuators such as bimorphs without any component optimization at low fields.

In this work, we tested the hypothesis that intrinsic kinematic constraints such as Donders' law are adopted by the brain to solve the redundancy in pointing at targets with the wrist. Ten healthy subjects were asked to point at visual targets displayed on a monitor with the three dof of the wrist. Three-dimensional rotation vectors were derived from the orientation of the wrist acquired during the execution of the motor task and numerically fitted to a quadratic surface to test Donders' law. The thickness of the Donders' surfaces, i.e., the deviation from the best fitting surface, ranged between 1 degree and 2 degrees, for angular excursions from +/-15 degrees to +/-30 degrees. The results support the hypothesis under test, in particular: (a) Two-dimensional thick surfaces may represent a constraint for wrist kinematics, and (b) inter-subject differences in motor strategies can be appreciated in terms of curvature of the Donders' surfaces.

Ideally, robots used for motor rehabilitation, in particular, during assessment, should minimally perturb the voluntary movements of a subject. In this paper, we show how a state-of-theart back-drivable robot, i.e., a robot that can be moved by the user with a low perceived mechanical impedance, when used for assessment can still perturb the voluntary movements of a subject. In particular, we show that, despite its low mechanical impedance, a robot may still not comply with the intrinsic kinematic constraints, which are of neural origin and are adopted by the human brain to solve redundancy in motor tasks. Specifically, the redundant task under consideration is the 2-D pointing task, which is performed by a subject with the sole use of the wrist [3 degree of freedom (DOF) kinematics]. Wrist orientations during pointing tasks are assessed in two different scenarios. In the first experiment, a lightweight handheld device is used, which introduces no loading effect. In the second experiment, similar pointing tasks are performed with the subject interacting with a state-of-the-art robot for wrist rehabilitation. In the first case, intrinsic kinematic constraints arise as 2-D surfaces embedded in the 3-D space of wrist configuration. Such surfaces are typically subject-dependent and reveal personal motor strategies. In the second case, a strong influence of the robot is remarked. In particular, 2-D surfaces still arise but are similar for all subjects and are referable to a mechanical origin (excessive loading by the robot). The assessment approach described in this paper, including both the experimental apparatus and dataanalysis method, can be used as a test for the degree of backdrivability of mechanisms and robots in relation to constraints of neural origin, thus allowing the design of robots that can actually cope with such constraints. The clinical potential impact is also discussed.

No abstract