We present and evaluate a pointing-gesture-based, bilateral human–robot interaction (HRI) with the humanoid robot Digit. Recently, humanoid robots have become more powerful and available, but they are still not helpful and accessible to non-technical users without programming experience. Therefore, we propose a pointing gesture-based interaction modality based on monocular video input. The pointing gestures are extracted using a pre-trained pose estimation model and, together with a verbal dialogue, serve as an exemplar interaction behavior for the task of moving heavy objects using the Digit. The interaction is evaluated quantitatively and qualitatively in a user study. The results of the study show promising results in terms of usability and implementation challenges for further research in this area.

This paper explores the adaptation of pseudo-rigid-body models (PRBMs) for simulating large geometric nonlinear deflections in passive exoskeletons, expanding upon their traditional application in small compliant systems. Utilizing the AnyBody modeling system, this study employs force-dependent kinematics to reverse the conventional simulation process, enabling the calculation of forces from the deformation of PRBMs. A novel approach, termed “Constraint Force”, is introduced to facilitate this computation. The approach is thoroughly validated through comparative analysis with laboratory trials involving a beam under bending loads. To demonstrate the functionality, the final segment of this study conducts a biomechanical simulation incorporating motion capture data from a lifting test, employing a novel passive exoskeleton equipped with flexible spring elements. The approach is meticulously described to enable easy adaptation, with an example code for practical application. The findings present a user-friendly and visually appealing simulation solution capable of effectively modeling complex mechanical load cases. However, the validation process highlights significant systematic errors in the direction and amplitude of the calculated forces (20% and 35%, respectively, in the worst loading case) compared to the laboratory results. These discrepancies emphasize the inherent accuracy challenges of the “Constraint Force” approach, pointing to areas for ongoing research and enhancement of PRBM methods.

In this study, we introduce a control strategy that combines Proportional-Derivative (PD) control with Iterative Learning Control (ILC) to enhance legged robot velocity control with only the inverse kinematics and no additional system identification. This approach leverages the realtime feedback capabilities of PD control for gait tracking while incorporating ILC's learning abilities to eliminate inaccuracies from unmodeled dynamics iteratively and to reach desired velocities without residual errors. By uniting these techniques, the proposed method empowers legged robots to adapt and optimize their control behavior, achieving and maintaining desired walking velocities. Experimental results on the physical legged robot Go1 demonstrate the effectiveness of the proposed approach, highlighting its adaptability and reliability in real-world scenarios. This research represents a first step towards overcoming high computational effort and extensive data collection for quadruped robot velocity tracking through onboard learning.

This letter introduces a control framework that leverages Lagrangian neural network (LNN) for computed torque control (CTC) of robotic systems with unknown dynamics. Unlike prior LNN-based controllers that are placed outside the feedback-linearization framework (e.g., feedforward), we embed an LNN inverse-dynamics model within a CTC loop, thereby shaping the closed-loop error dynamics. This strategy, referred to as LNN-CTC, ensures a physically consistent model and improves extrapolation, requiring neither prior model knowledge nor extensive training data. The approach is experimentally validated on a robotic arm with four degrees of freedom and compared with conventional model-based CTC, physics-informed neural network (PINN)-CTC, deep neural network (DNN)-CTC, an LNN-based feedforward controller, and a PID controller. Results demonstrate that LNN-CTC significantly outperforms model-based baselines by up to $30 \,\%$ in tracking accuracy, achieving high performance with minimal training data. In addition, LNN-CTC outperforms all other evaluated baselines in both tracking accuracy and data efficiency, attaining lower joint-space RMSE for the same training data. The findings highlight the potential of physics-informed neural architectures to generalize robustly across various operating conditions and contribute to narrowing the performance gap between learned and classical control strategies.

This work presents a body velocity control strategy for quadruped robots. Such control typically requires accurate kinematic and dynamic model knowledge, which is very challenging because of the multidimensional input-output system and the ground contact. Based on the inverse kinematics, we propose a Proportional-Derivative controlled robot that uses Iterative Learning Control to learn discrete body velocities, which are then generalized using the Gaussian Process Regression model for each joint separately. This controller design enables onboard control and learning in real-time without any simulation. This study illustrates the effectiveness of the proposed methodology over a range of velocities while emphasizing the minimal computational effort associated with its application in a practical context.