Tactile sensation is one of the most important sensing methods of humans, which helps humans complete complex decisions and actions by itself or in cooperation with other sensations. Similarly, tactile sensors are also crucial for robots to interact with the environment and fulfill dexterous object manipulation. In this study, we present DotView, a compact, low-cost, and high-performance tactile sensor that utilizes a biomimetic micro-structured soft layer to generate rich, high-resolution tactile images captured by a commercialized capacitive sensing array, which exhibits sensitive and robust responses under pressure, shear and torsion applied by different indenters. Furthermore, we utilize a neural network to estimate multiaxial contact forces and torques, and achieve good accuracy in different contact geometries. Finally, we demonstrate the capability of DotView in a dexterous manipulation scenario by performing the task of guiding a tool through a maze with only tactile feedback.

Designing anthropomorphic prosthetic hands that approach human-level performance remains a great challenge. Commercial prosthetics are still inferior to human hands in several important properties, such as weight, size, fingertip force, grasp velocity, and active and passive dexterities. We present a novel design based on the under-actuated da Vinci’s mechanism driven by a flexible twisted string actuator (TSA). The distributed drive scheme allows structural optimization using a motion capture database to reproduce the natural movement of human hands while keeping adaptability to free-form objects. The application of TSA realizes a high conversion from motor torque to tendon contraction force while keeping the structure light, flexible, and compact. Our anthropomorphic prosthetic hand, consisting of six active and 15 passive degrees of freedom, has a weight of 280 g, approximately 70% of that of a human hand. It passed 30 of the 33 Feix grasp tests on objects in daily living and retained a loading capacity of 5 kg. This simple but intelligent mechanism leads to excellent stability and adaptability and renders feasible wide applications in prosthetics and in service robots.

Dexterous hands play vital roles in tasks performed by humanoid robots. For the first time, we quantify the correlation between design variables and the performance of 65 dexterous hands using Cramér’s V. Comprehensive cross-correlation analysis quantitatively reveals how the performance, such as speed, weight, fingertip force, and compactness are related to the design variables including degrees of freedom (DOF), structural form, driving form, and transmission mode. This study shows how various design parameters are coupled inherently, leading to compromise in performance metrics. These findings provide a theoretical basis for the design of dexterous hands in various application scenarios and offer new insights for performance optimization.

Tactile sensing technologies enable robots to interact with the environment in increasingly nuanced and dexterous ways. A significant gap in this domain is the absence of curved tactile sensors, which are essential for performing sophisticated manipulation tasks. In this study, we present DotTip, a tactile fingertip featuring a three-dimensional curved perceptual surface that closely mimics human fingertip morphology. A convolutional neural network-based deep learning framework precisely calculates the contact angles and forces from the sensor tactile images, achieving mean errors of 1.56 ${}^{\circ }$ and 0.28 N, respectively. DotTip's performance is evaluated in real-world tasks, demonstrating its efficacy in tactile servoing, slip prevention, and grasping, along with the more challenging benchmark task of controlling a joystick. These findings demonstrate that DotTip possesses superior 3D tactile sensing capabilities necessary for fine-grained dexterous manipulations compared to its flat counterparts.

Robotic racket sports provide exceptional benchmarks for evaluating dynamic motion control capabilities in robots. Due to the highly non-linear dynamics of the shuttlecock, the stringent demands on robots' dynamic responses, and the convergence difficulties caused by sparse rewards in reinforcement learning, badminton strikes remain a formidable challenge for robot systems. To address these issues, this study proposes DTG-IRRL, a novel learning framework for badminton strikes that integrates imitation-relaxation reinforcement learning with dynamic trajectory generation. The framework demonstrates significantly improved training efficiency and performance, achieving faster convergence and twice the landing accuracy. Analysis of the reward function within a specific parameter space hyperplane intuitively reveals the convergence difficulties arising from the inherent sparsity of rewards in racket sports and demonstrates the framework's effectiveness in mitigating local and slow convergence. Implemented on hardware with zero-shot transfer, the framework achieves a 90% hitting rate and a 70% landing accuracy, enabling sustained humanrobot rallies. Cross-platform validation using the UR5 robot demonstrates the framework's generalizability while highlighting the requirement for high dynamic performance of robotic arms in racket sports.