force-torque sensing

To address the urgent need for compact, lightweight six-axis force/torque sensors in legged robots, this work proposes a non-contact sensing architecture based on optocouplers—overcoming key limitations of conventional strain-gauge sensors, such as mechanical fragility and significant thermal drift. A differentiable multiphysics parametric–performance mapping model is developed to enable joint global optimization of sensitivity and measurement error. The resulting sensor weighs less than 85 g and features full-scale ranges of ±150 N and ±5 N·m. Calibration and dynamic testing demonstrate a static error of ≤1.2%, with excellent agreement between theoretical predictions and empirical measurements. Integrated successfully onto a quadrupedal robot, the sensor accurately captures high-frequency ground–foot interaction forces. This work establishes a new paradigm for high-precision, robust force perception in highly dynamic legged platforms.

This work addresses the challenge of detecting and precisely localizing foot-ground contact points for quadrupedal robots. We propose a contact perception method integrating distributed joint strain-based torque sensors with hip-mounted six-axis force/torque sensors. Leveraging a generalized momentum observer framework, our approach achieves real-time estimation of contact forces and positions via multi-sensor fusion, high-accuracy calibration, and a physics-driven estimation strategy that obviates motor current modeling and friction compensation. Key contributions include: (i) the design of low-cost, highly linear distributed strain-based torque sensors; and (ii) a lightweight perception architecture that requires no dynamics modeling or parameter identification. Simulation and experimental results demonstrate a contact force estimation error below 0.2 N, sub-centimeter contact point localization accuracy (mean error: 0.8 cm), and sensor measurement accuracy of 96.4%.

Traditional strain-based six-axis force/torque (F/T) sensors suffer from contact-dependent measurement, reliance on external amplifiers and data acquisition (DAQ) systems, and inherent trade-offs between compactness and high accuracy. To address these limitations, this paper proposes a fully integrated, contactless inductive F/T sensor. Force information is acquired via displacement sensing of a conductive target, while a CAN-FD signal processing module is embedded directly onto the PCB, enabling 4 kHz high-speed sampling and on-chip closed-loop processing. We introduce a novel rational function modeling approach to significantly enhance system linearity and calibration accuracy. The sensor achieves crosstalk < 0.5%, force resolution of 0.03 N, and over 55,000 quantization levels. Static evaluation shows superior RMSE and R² compared to state-of-the-art nonlinear models. Overall performance exceeds that of commercial counterparts, fulfilling the stringent requirements of precision robotics for high-accuracy, miniaturized, and minimally intrusive integrated sensing.

This study addresses the insufficient accuracy of joint torque estimation in collaborative robots within low-torque regimes, primarily caused by gearbox static friction and nonlinear current-to-torque relationships. To overcome this limitation, the authors propose a non-contact optical joint torque sensor based on micro-deformation detection of an elastic structure. The design incorporates a quadrature-redundant array of optical reflective sensors to enhance sensitivity and signal-to-noise ratio, complemented by quadratic programming-based calibration, temperature drift compensation, and rational function fitting. Experimental results demonstrate exceptional performance, with a full-scale maximum error of only 0.083%, an RMS error of 0.0266 Nm, and a 3σ resolution of 0.0224 Nm at 1 kHz—representing a 2.14-fold improvement over conventional least-squares methods. The system exhibits superior precision and robustness to disturbances in low-torque tracking and admittance control tasks.

This study addresses the limitations of traditional human dynamics analysis, which relies on contact-based force/torque sensors and controlled environments, rendering it unsuitable for non-contact scenarios. To overcome this, the authors propose an optics-mechanics integrated framework for simultaneous kinematic and dynamic estimation. By formulating a constrained multibody dynamics model, the method leverages vision-based kinematic measurements as non-contact inputs and employs a genetic algorithm to optimize joint torque identification. This approach represents the first demonstration of non-contact dynamic parameter estimation for multibody systems using only visual data and a mechanical model, eliminating dependence on force sensors. Experimental validation on an air-bearing platform shows a mean absolute error of 0.46 Nm in wrist joint torque estimation and a forward-predicted angular velocity error as low as 0.006 rad/s.

This work addresses the lack of embeddable miniature multi-axis force/torque sensors in tweezer-like tools, which hinders real-time force feedback and failure detection during delicate manipulation. The authors propose PinFT—a compact five-axis capacitive force/torque sensor designed for integration into both tips of a pair of tweezers. Leveraging a three-layer PCB stack, segmented plated-through-hole electrodes, and a silicone dielectric layer, PinFT transduces five-degree-of-freedom loads via 3D displacement of a central stainless-steel pin, with measurement accuracy enhanced through high-order polynomial calibration. Validated on 3D-printed tweezers and a parallel-jaw gripper platform, the system achieves mean absolute errors of 0.23 N in force and 2.5 mN·m in torque (R² > 0.97), successfully enabling tasks such as sub-millimeter component handling, simulated hair plucking, and soft material tearing while reliably detecting slip and ejection failures to significantly enhance micromanipulation perception.

This study addresses the lack of high-precision, easily integrable distal force-sensing solutions in cable-driven robotic instruments for minimally invasive surgery, a limitation that hinders the development of haptic feedback and force-controlled autonomous operations. The authors propose a novel six-axis force sensor integration design at the instrument’s distal end that requires no specialized fabrication equipment and, for the first time, employs a Transformer neural network to fuse multimodal sensor and robot state data, effectively compensating for internal force disturbances induced by cable actuation. Without compromising the instrument’s original functionality, the approach achieves high-accuracy distal force estimation with normalized errors below 6%, significantly outperforming purely proximal data-driven methods under unseen operating conditions. This work establishes a reproducible technical pathway toward skill assessment and force-aware autonomous surgical systems.

This study addresses the challenge of distorted haptic feedback in surgical robot training caused by non-contact disturbances—such as gravity, sensor bias, and mounting misalignment—affecting wrist-mounted force/torque sensors. To overcome this, the authors integrate a wrist sensor into the low-cost RoboScope platform and propose a real-time adaptive compensation method based on recursive least squares (RLS). This approach dynamically eliminates interference without requiring pre-collected data or repeated calibration, achieving fully online non-contact force compensation for the first time. Experimental results demonstrate that the method reduces force and torque errors by over 95% and 91%, respectively, significantly outperforming existing techniques. The proposed solution thus provides reliable, high-fidelity force perception essential for cost-effective, high-precision haptic surgical training.

This study addresses the ambiguity and inaccuracies in existing definitions of interaction forces and internal loads in redundantly actuated parallel mechanisms, which have led to erroneous force analyses and control deviations. To resolve this issue, the work rigorously clarifies these two concepts and proposes a unified dynamic modeling framework based on null-space torque decomposition and joint torque vector synthesis. The resulting formulation provides a clear and unambiguous analytical foundation specifically tailored for such mechanisms. Validation through representative case studies demonstrates that the proposed method effectively corrects erroneous results reported in the literature and significantly enhances force control accuracy in both balancing and manipulation tasks for redundantly actuated parallel mechanisms.

This study addresses the longstanding lack of effective haptic feedback in surgical robots and the challenge of acquiring high-quality training data under representative external forces without adding sensors. To this end, the authors propose a non-invasive, six-degree-of-freedom parallel motor-cable system arranged around the workspace of the RAVEN II surgical robot. By applying precisely controlled cable tensions to the robot’s end-effector, the system delivers accurate external forces without impeding its motion. The framework integrates custom motor hardware, a tension control algorithm, sensor drivers, and a simulation module, enabling—for the first time—high-precision force application using a parallel cable-driven architecture. Experimental results demonstrate a force control error of less than 1 N, providing high-fidelity training data for force-perception learning in robotic surgery.