Inferensys

Glossary

Force/Torque Sensing

Force/torque sensing is the measurement of the forces and torques applied at a robot's wrist or end-effector, enabling compliant control and precise manipulation.
Control room desk with laptops and a large orchestration network display.
ROBOT MANIPULATION AND GRASPING

What is Force/Torque Sensing?

Force/torque sensing is the precise measurement of multi-axis contact forces and rotational moments applied at a robot's wrist or end-effector.

Force/torque (F/T) sensing is the direct measurement of the six-dimensional wrench—three linear forces (Fx, Fy, Fz) and three rotational torques (Tx, Ty, Tz)—applied at a robot's wrist or end-effector. This is achieved using a six-axis force/torque sensor, a specialized transducer typically based on strain gauge technology mounted on a deforming structure. The sensor provides a real-time, high-fidelity signal of external contact loads, enabling robots to transition from purely position-controlled machines to systems that can feel and respond to physical interaction.

This sensory feedback is foundational for compliant control strategies like impedance control and admittance control, which allow a robot to modulate its stiffness and motion in response to contact forces. In manipulation, it enables precise tasks such as compliant assembly (e.g., peg-in-hole), delicate object handling, and surface following by closing the loop between physical interaction and motion command. It is a critical enabler for collaborative robots (cobots), providing the data necessary for safe human-robot interaction and contact detection.

ROBOT MANIPULATION AND GRASPING

Key Applications of Force/Torque Sensing

Force/torque sensing enables robots to interact intelligently with their environment by measuring contact forces. This capability is foundational for tasks requiring precision, safety, and adaptability.

01

Compliant Control and Assembly

Force/torque sensors enable compliant control strategies like impedance control and admittance control. These allow a robot to behave like a spring-damper system, yielding to contact forces rather than fighting them. This is critical for delicate assembly tasks (e.g., inserting a peg into a hole, screwing components) where geometric uncertainty exists. The sensor provides real-time feedback, allowing the controller to adjust the robot's pose to accommodate misalignments and prevent jamming or part damage.

02

Grasp Force Regulation

Precise control of grip strength is essential for handling a wide variety of objects. A force/torque sensor at the wrist or within a dexterous gripper measures the contact forces during a grasp. This feedback allows the robot to:

  • Apply the minimum necessary force to securely lift fragile items (eggs, electronic components).
  • Detect slip by monitoring shear forces and increase grip to prevent dropping.
  • Achieve form closure or force closure by actively adjusting finger positions based on tactile feedback, ensuring stable manipulation.
03

Human-Robot Collaboration (HRC)

In shared workspaces with collaborative robots (cobots), force/torque sensing is a primary safety and interaction modality. It enables:

  • Physical guidance programming: An operator can directly move the robot arm by applying force, and the sensor records the motion for playback.
  • Reactive collision detection: Unexpected contact with a human or obstacle generates a force/torque signature that triggers an immediate protective stop.
  • Hand-over tasks: The robot can sense when a human takes an object from its gripper and release its grasp accordingly.
04

Contact State Estimation and Haptic Exploration

Beyond simple force measurement, these sensors allow a robot to infer the contact state with its environment. By analyzing the wrench (combined force and torque vector), the system can determine:

  • If contact is point, line, or surface contact.
  • The center of pressure on an end-effector.
  • When a part has been seated correctly in a fixture. This information is vital for haptic exploration, where a robot actively probes an object or surface to identify its properties (e.g., stiffness, texture, edges) or locate features by touch.
05

Payload Identification and Tool Center Point (TCP) Calibration

A force/torque sensor can automatically identify the inertial properties of an unknown payload attached to the end-effector. By executing specific excitation motions and measuring the dynamic forces, the robot can calculate the payload's mass, center of mass, and inertia matrix. This is crucial for dynamic model-based controllers like Model Predictive Control (MPC). Similarly, the sensor is used to precisely calibrate the Tool Center Point (TCP) by touching a fixed point in space from different orientations and calculating the offset.

06

Deburring, Polishing, and Surface Finishing

These material removal tasks require maintaining a constant normal force against a contoured surface. A force/torque sensor provides closed-loop feedback to adjust the robot's trajectory in real time, compensating for part geometric variations and tool wear. This ensures a uniform finish quality. Applications include:

  • Robotic deburring of machined metal parts.
  • Polishing complex molds or automotive body panels.
  • Sanding wooden components. The alternative—rigidly programming the path—often leads to inconsistent results or tool damage.
CONTROL ARCHITECTURE

Force-Based Control Strategies: Impedance vs. Admittance

A comparison of the two primary paradigms for implementing compliant robotic control using force/torque sensor feedback.

Core Feature / MetricImpedance ControlAdmittance Control

Control Law

Force = Impedance * (Velocity Error)

Velocity = Admittance * (Force Error)

Primary Input

Desired end-effector motion trajectory

Measured external contact force/torque

Primary Output

Commanded joint torques

Desired corrective motion (position/velocity)

Inner Control Loop

Direct torque control

High-gain position/velocity control

Stability in Hard Contact

Inherently stable; behaves like a passive mechanical system

Can become unstable; requires careful tuning of the outer loop

Transparency (Free Motion)

Lower; feels 'mushy' or damped due to simulated dynamics

Higher; robot feels rigid and responsive when not in contact

Implementation Hardware

Requires high-fidelity joint torque control (e.g., series elastic actuators, torque sensors)

Requires a high-precision, high-bandwidth position/velocity controller and an external F/T sensor

Typical Applications

Physical human-robot interaction (pHRI), tasks requiring gentle contact (polishing, wiping)

Precision assembly (peg-in-hole), heavy payload manipulation, industrial grinding

FORCE/TORQUE SENSING

Frequently Asked Questions

Force/torque sensing is a foundational technology in advanced robotics, enabling machines to feel and respond to physical contact. This FAQ addresses the core technical questions about how these sensors work, their applications, and their integration into robotic control systems.

A force/torque (F/T) sensor is a transducer that measures the three orthogonal force components (Fx, Fy, Fz) and three orthogonal torque components (Tx, Ty, Tz) applied at a single point, typically mounted between a robot's wrist and its end-effector. It works by using an array of strain gauges bonded to a precisely machined, elastic structure (often called a transducer body). When forces and torques are applied, the structure deforms minutely, causing a change in the electrical resistance of the strain gauges. These resistance changes are measured via a Wheatstone bridge configuration, amplified, and converted via an analog-to-digital converter into a six-degree-of-freedom (6-DoF) wrench vector that a robot's controller can interpret.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.