Whole-Body Sensory Concept for Compliant Mobile Robots

Unfreezing Social Navigation: Dynamical Systems based Compliance for Contact Control in Robot Navigation

“…Herewith, mitigating the impact during the transient phase. Unlike the work in [19] where a stiff hull was developed mounted on an FT sensor. However, this brings a challenge in accurately estimating contacts on the surface of the hull, which we have tackled by learning the non-linearity of the passive-compliance through data of known impact forces.…”

“…A similar principle for closed-loop force control with 6-axes Force/Torque was offered in [19] with a focus on learning touch commands on a stiff hull. While the method in [20] offered an impedance control response when guiding a person's motion while limiting the maximum guidance force.…”

Force-Torque Sensor Disturbance Observer Using Deep Learning

“…In order to estimate these forces precisely, the ten inertial parameters of the load should be known, namely: m, c and the values of I. In the literature, researchers usually use identification methods to obtain these values and apply equations (4) and (5) to calculate the non-contact forces and torques. However, the accuracy of estimating non-contact forces and torques based on identification is dependent on the accuracy of the center of mass position of the load c and the calculation of the kinematic vectors α, ω andω in the same frame.…”

“…One of the first works in robotics to use neural networks for estimating forces, was described in [4], where a feedforward neural network approximated two-dimensional forces based on the robot's joint positions, velocities and accelerations. More recently, Kollmitz et al estimated a six-dimensional contact force on a robot platform using a timedelay neural network [5], where the network delayed the inputs (wrench and acceleration measurements) in order to include temporal information. Another network architecture that makes use of sequential data are recurrent neural networks (RNNs).…”

Cognitive and motor compliance in intentional human-robot interaction

“…mediated by physical objects [1], and behavior improvement [2]), leaving aside the social dimension of the interaction with the robot partner. With some exceptions, direct contact in autonomous behavior has been avoided [3]. Moreover, in social robotics, the non-verbal aspects of interaction have been studied from diverse modalities (e.g.…”