Underactuation has become, in recent years, more and more prevalent in robotic fingers since it provides the latter with the ability to mechanically adapt to the shape of the objects seized. To improve the usually simplistic control schemes of these fingers, and possibly to provide force feedback or control, tactile sensors are typically used. However, another promising avenue, as presented in this paper, is rather to use information provided by proprioceptive (i.e., internal) sensors. Most interestingly, this can be done using only the torque and position sensors typically found attached to the actuator(s) of these fingers. Because a relationship exists between the stiffness of an underactuated finger as seen from its actuator and the contact locations on its phalanges, it is possible to estimate one from the other. In this paper, a proprioceptive tactile sensing algorithm based on this technique is presented. It is concluded that within certain theoretical and practical limits, it is possible to extract tactile data from a self-adaptive finger, namely position and magnitude of the contact forces, without actually using any physical tactile sensors.

The authors report on the analysis and experimental evaluation of a parallel Schönflies-motion generator (SMG) intended for fast pick-and-place operations and actuated with cylindrical drives (C-drives). Its CRRHHRRC isostatic architecture offers high rotability of the moving platform and a reduced number of limbs, as compared to robots available on the market. A simulation model using the dynamics of the robot as well as an experimental prototype are described. The results obtained are used to assess the pros and cons of two alternative C-drive architectures. A workspace analysis including the feasible wrenches is conducted. The industry standard Adept test cycle is used to evaluate the performance of the prototype. The data obtained prove that C-drives augmented with strain-wave gears give the SMG significantly better wrench capabilities and lower trajectory-tracking error, compared to their alternative C-drives based on cables.

Trajectory planning and an efficient control scheme play a crucial role in improving the performance of pick-and-place robots. This paper introduces a novel method of trajectory planning with cycle time and path constraints. Assuming that a smooth trajectory is given, to be followed within a prescribed cycle time, the newly proposed method of trajectory planning removes the torque peaks of the actuators by a suitable scheduling of the velocity of the moving plate. Since pick-and-place robots are usually expected to meet the end poses in a certain time span, while disregarding the intermediate poses, the velocity can be tuned properly around the critical points of the trajectory by means of a time-scaling function. Moreover, the authors report the formulation of a linear quadratic regulator (LQR) controller with normalized variables to be used in conjunction with our trajectory-tracking control scheme for an in-house-developed Schönflies-motion generator. This parallel robot offers a functionally symmetric, single-loop architecture, with an isostatic kinematic chain, and virtually unlimited rotatability of its gripper. A comparison between two actuation systems developed by the authors is conducted via simulation results.