CoinFT: A Coin-Sized, Capacitive 6-Axis Force Torque Sensor for Robotic Applications

arXiv:2503.19225 · cs.RO, cs.HC · Submitted 2025-03-25 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa, and with me are Dev and Taro, guest researcher.

Dev: Today's paper: "CoinFT: A Coin-Sized, Capacitive 6-Axis Force Torque Sensor for Robotic Applications".

Rosa: CoinFT introduces a compact, light, and low-cost capacitive 6-axis force/torque (F/T) sensor designed for various robotic applications.

Dev: First, who's behind it and why it matters.

Paper summary: Rosa: To break down this paper further, they are introducing CoinFT as a capacitive six-axis force/torque sensor that is designed to be compact, light, low-cost, and robust. The core claim is that this specific design allows for contact-rich robot interactions in domains such as drones and wearable haptic devices.

Dev: It’s not just about being small; the paper points out their performance metrics too, mentioning an average root-mean-squared error of zero point one six N for force and one point zero eight mNm for moment when the input ranges from zero to fourteen N and zero to five N in normal and shear directions, respectively.

Taro: Those specific numbers give us a good baseline for what we can expect from this sensor when deployed in real-world scenarios, which is important for planning complex behaviors.

Rosa: That level of detail on the expected error range makes it tangible; it shows how accurate this low-cost device is supposed to be in practice.

Dev: And they mention that the microcontroller interrogates the electrodes in different subsets to improve sensitivity for measuring those six axes of force and torque.

Taro: That aspect about using different electrode configurations sounds like a clever way to boost the sensing capability without necessarily increasing the physical size of the sensor itself, which is something we always look for in system design.

Conclusion: Rosa: Looking at the title "CoinFT: A Coin-Sized, Capacitive six-Axis Force Torque Sensor for Robotic Applications," it really captures the essence of what they are presenting—a sensor focused on small size and multi-axis measurement capability. The authors, including Hojung Choi, Jun En Low, Tae Myung Huh, Seongheon Hong, Gabriela A. Uribe, Kenneth A. W. Hoffmann, Julia Di, Tony G. Chen, Andrew A. Stanley and Mark R. Cutkosky are clearly a strong team tackling this problem from various angles like field robotics and control engineering.

Dev: I think the real implication here is taking force sensing out of the realm of expensive or bulky hardware and making it something accessible for a wider range of robotic platforms, which is crucial for scaling up autonomous systems.

Taro: For autonomy, having a reliable, low-cost way to sense interaction forces could mean that smaller robots can learn and adapt to physical environments much faster than they currently can.

Rosa: That’s right; the ability to use this sensor in things like wearable haptics opens up new ways for humans to interact with technology in a more nuanced, physical way.

Dev: And from an engineering standpoint, it suggests that we don't always have to rely on highly specialized or fragile sensing technologies when developing robot end-effectors or other contact-sensitive systems.

Rosa: So, to wrap up this discussion on CoinFT: A Coin-Sized, Capacitive six-Axis Force Torque Sensor for Robotic Applications, the paper shows a design that balances physical constraints with necessary sensing performance for diverse robotic tasks.

Dev: It really highlights how capacitive sensing can be applied effectively when you need both a small package and multi-axis force torque measurement.

Taro: The potential impact is in enabling more flexible and adaptable robotic systems across various embodiments, from drones to personal haptic gear, because of this compact sensor.

Hojung Choi, Jun En Low, Tae Myung Huh, Seongheon Hong, Gabriela A. Uribe, Kenneth A. W. Hoffmann, Julia Di1, Tony G. Chen

Stanford University, CA, USA. · University of California Santa Cruz, CA, USA. · Reality Labs Research, Meta Platforms Inc., WA, USA

cs.RO, cs.HC

Submitted: 2025-03-25

Updated: 2026-09-27

Project page: https://coin-ft.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 77/100

The gist: CoinFT introduces a compact, light, and low-cost capacitive 6-axis force/torque (F/T) sensor designed for various robotic applications.

Key concepts

Capacitive Sensing: The sensor measures forces and torques by detecting changes in electrical capacitance between two layers of PCBs. This works by manipulating the distance between electrodes; when a physical load is applied, it changes the gap, altering the capacitance signal. The microcontroller interprets these specific signal patterns to determine force and torque values.
Dual-Mode Electrode Switching: CoinFT achieves 6-axis sensing by switching between two electrode configurations: 'normal mode' and 'shear mode'. In normal mode, it is sensitive to inputs like Fz, Mx, and My. In shear mode, it focuses on inputs such as Fx, Fy, and Tz. This firmware-controlled switching allows the sensor to measure all six degrees of freedom effectively.
Force/Torque Measurement: The sensor quantifies physical interactions by measuring force (linear) and torque (rotational). It uses a specific combination of electrode arrangements in either mode to distinguish between different directions of applied loads. This allows the system to accurately determine how much push or pull is being exerted, as well as rotational twisting forces.
Robustness Against Impact: CoinFT demonstrates reliability when subjected to sudden mechanical shocks. It can accurately measure small loads (1 N, 2 N, 3 N) even after sustaining a significant impact that would normally cause a much larger reading (like 180 N). This robustness is crucial for real-world robotic interactions where unexpected impacts are common.

Terminology

Robustness — 16× in this paper · explained in 5 episodes
Calibration — 15× in this paper · explained in 7 episodes
Alignment — 10× in this paper · explained in 1 episode
Accuracy — 9× in this paper · explained in 1 episode
Fiber Bragg Grating — 2× in this paper
Hysteresis — 2× in this paper · explained in 3 episodes
PID Controller — 2× in this paper · explained in 2 episodes
Signal-to-noise ratio (SNR) — 2× in this paper · explained in 2 episodes
Ablation — 1× in this paper · explained in 1 episode
Consistency — 1× in this paper · explained in 3 episodes
Core — 1× in this paper · explained in 2 episodes
Decoupling — 1× in this paper · explained in 2 episodes
Foundation Model — 1× in this paper · explained in 3 episodes
Foundation Models — 1× in this paper · explained in 7 episodes
Physical Constraints — 1× in this paper · explained in 7 episodes
Redundancy — 1× in this paper · explained in 2 episodes
Structural Integrity — 1× in this paper · explained in 3 episodes

Summary

CoinFT introduces a compact, light, and low-cost capacitive 6-axis force/torque (F/T) sensor designed for various robotic applications. The sensor's combination of features enables contact-rich robot interactions across domains including drones, robot end-effectors, and wearable haptic devices.

The gist: CoinFT is a capacitive 6-axis F/T sensor that is compact, light, low-cost, and robust with an average root-mean-squared error of 0.16 N for force and 1.08 mNm for moment when the input ranges from 0∼14 N and 0∼5 N in normal and shear directions, respectively.

CoinFT Design

CoinFT is a stack of two rigid PCBs with comb-shaped electrodes connected by an array of silicone rubber pillars. The sensor has a circular sensing area of 20 mm in diameter and approximately 2 mm in thickness, weighing 2 g. Each PCB is multilayered, featuring four quadrants of a pair of comb-shaped electrodes on one side and a plane electrode for passive shielding on the other side. The design leverages the microcontroller's ability to switch between two different electrode configurations—normal mode and shear mode—to achieve multi-axis sensing.

Working Principle

CoinFT measures 6-axis force and torque by observing changes in capacitance signal patterns between the upper sensing layer and the lower sensing layer. The microcontroller firmware is programmed to switch the pair of rigid PCBs between two different electrode configurations every sampling sequence: normal mode and shear mode. In normal mode, it is sensitive to inputs that change distance, such as Fz, Mx, and My. In shear mode, it is sensitive primarily to shear inputs such as Fx, Fy, and Tz. The comb-shaped electrodes are populated in a way that allows the first and third quadrant of the shear mode to be sensitive to loads in the X direction while the second and fourth quadrants are sensitive to the Y direction.

Sensor Fabrication

The fabrication process involves several steps:

A 127 um polyimide film pillar mask is cut by a UV laser cutter.
Vacuum degassed uncured silicone rubber is smeared on the mask to fill the cavities for pillars.
The lower sensing layer PCB is primed and aligned onto the pillar mask, which is stacked with an acrylic plate and a 3.1 kg mass to minimize base layer thickness during casting.
The assembly is cast inside a pressure chamber for at least 12 hours at approximately 414 kPa to compress any remaining micro-bubbles in the silicone rubber.
The upper sensing layer PCB is primed, and fresh uncured silicone is spread and spin-coated to achieve a uniform and thin layer.
The pillar layer is attached to the upper sensing layer PCB with a 203 um polyimide film spacer between them, controlling the distance between the two PCBs.
The top shield layer fPCB is adhered to the other side of the upper sensing layer PCB using an adhesive (Loctite 401 Instant Adhesive).

Sensor Characterization and Performance

Quantitative analysis using finite element analysis (FEA) and real sample testing was conducted. Key characterization results include:

: The average root-mean-squared error is 0.16 N for force and 1.08 mNm for moment when the input ranges from 0∼14 N and 0∼5 N in normal and shear directions, respectively.

: The minimum detectable force of CoinFT is 20 mN.

The dual-mode electrode-switching strategy produces a distinguishable signal pattern from its 12 sensing electrodes under different force and torque loads. The mechanical bandwidth of CoinFT is approximately 97 Hz in the shear direction. Furthermore, CoinFT demonstrates reliable robustness against impacts, accurately measuring small loads of 1 N, 2 N, and 3 N even after sustaining an impact from a hammer that causes a force reading of 180 N.

Applications

CoinFT is validated through two representative applications:

A multi-axial contact-probing experiment where a CoinFT mounted beneath a hemispherical fingertip measures 6-axes of force and torque representative of manipulation scenarios.
An attitude-based force-control task on a drone, demonstrating the use of CoinFT for contact modulation through PID force control and the deployment of objects onto environmental surfaces by applying a controlled range of force.

The sensor's versatility extends to wearable haptic devices, where it measures and controls contact force between the device and the human subject. It has also been attached to robot fingertips for learning force-informed actions from kinesthetic demonstrations, allowing it to sense contact force for controlled, forceful interaction or track the contact point via force and torque information.

Improvements for AI systems

Here are specific improvements to AI systems based on the capabilities demonstrated by CoinFT, and what those improved systems could achieve:

A more robust and cost-effective perception module for robotic manipulation tasks (e.g., grasping, assembly).
An enhanced force/torque feedback loop for dexterous robotic hands or end-effectors that requires high sensitivity in fine manipulation scenarios (e.g., palpating soft tissue, precise insertion).
A low-cost, impact-tolerant sensing system for autonomous aerial vehicles (drones) operating in unstructured environments, allowing for force-controlled contact modulation during perching or surface attachment without requiring expensive, fragile commercial sensors.
A perception system that allows robots to infer the presence and location of contact surfaces by analyzing subtle force/torque signatures (contact-probing) rather than relying solely on visual cues or pre-programmed trajectories.
A wearable haptic device capable of providing highly accurate, multi-axial tactile feedback to humans, enabling more nuanced interaction with virtual or physical objects based on perceived normal vs. shear forces.

These improved AI systems can achieve the following specific functions:

The robot could perform delicate tasks like assembling small components or handling fragile objects (e.g., a coin-sized device) with micron-level force control, preventing damage that would occur with bulky or inaccurate sensors like the Gamma sensor mentioned in the paper.
A prosthetic hand AI could learn to feel textures and material compliance during interaction, enabling it to grasp soft biological materials (like tissue) safely by modulating force based on real-time shear and normal readings, which is critical for medical robotics.
An autonomous drone could reliably land on a complex or uneven surface without crashing by using CoinFT feedback to modulate its landing forces in real-time, allowing it to maintain stable contact during perching or when attaching payloads (like sensors) to unpredictable surfaces like tree branches.
A robotic system could autonomously navigate cluttered environments by feeling for edges and surfaces through contact probing, allowing it to perform complex manipulation without relying solely on high-cost visual mapping alone.
A wearable AI could provide users with highly intuitive haptic feedback that distinguishes between a gentle normal touch and a directional shear force (e.g., sliding an object), improving the fidelity of remote manipulation or teleoperation tasks involving physical objects.