3D Printed Low-Cost Force-Torque Sensors

Norman Hendrich, Florens Wasserfall, Jianwei ZhangIEEE Access 2020 · DOI

Abstract

Force sensing is essential for many manipulation tasks and, more generally, for all robots physically interacting with their environment. While multi-axis force/torque sensors are readily available commercially, their cost and complex integration have so far limited a wide deployment. In this paper, we introduce a modular approach to design and to integrate low-cost force sensors directly into 3D printed robot parts. Based on off-the-shelf optical sensors embedded into deformable structures, sensitivity and load capacity can be selected from a wide range. A working six-axis sensor, including electronics, can be built for less than 20 dollars, plus a few hours of 3D printing. We present tested example designs for sensors of different complexity, from a basic one-dimensional deflecting beam to six-axis sensors with custom shapes. We summarize the basic sensor layout geometries, explain key 3D printing and integration aspects, discuss sensor calibration, and describe our Arduino firmware and ROS-based drivers.

Introduction

In robotics, "feeling" the environment is as critical as seeing it. However, high-fidelity force-torque (F/T) sensors, typically based on strain gauges, often cost between $4,000 and $5,000. This high price, coupled with the need for specialized amplifiers and bulky shielded cabling, restricts their use in small-scale mobile robots or experimental grippers. Furthermore, 3D-printed materials are often avoided for precision sensing due to their anisotropic nature and susceptibility to "creep" or permanent deformation.

Hendrich et al. address these challenges by proposing a modular framework to create custom, multi-axis F/T sensors for under $20 using standard 3D printers and off-the-shelf optical sensors.

Overview of 3D-printed force-torque sensor prototypes, internal spring structures, and the low-cost optical components used (inset).

Methodology

The core of this approach is a modular design where a fixed sensor carrier is paired with various 3D-printed elastic structures.

Sensing Mechanism

Instead of measuring internal material strain, the system measures the displacement of a structure under load using inexpensive ($0.50–$1.00) optical sensors:

  • Reflex-type (Proximity): Measures distance to a reflecting surface.
  • Fork-type (Interrupter): Measures the degree to which a "fin" blocks a light beam.

Mechanical Design

The authors utilize OpenSCAD to design elastic members such as cantilevers and planar spiral springs. Spiral springs are particularly effective as they allow for deflection in all three axes (x, y, z) while remaining compact and printable without support structures.

Software-Defined Precision

Since 3D-printed plastic behavior is complex, the researchers rely on a data-driven calibration approach rather than mechanical perfection. They employ Taylor-series expansion (including linear and quadratic terms) to map raw sensor voltages to force/torque values. To handle the inherent non-linearities and cross-talk of 3D-printed plastic, they use:

  • LASSO Regression: To keep models sparse by identifying the most relevant sensors for each axis.
  • Huber Regression: To ensure calibration is robust against outliers.

Key Results and Prototypes

The modularity of the design allows researchers to adjust sensitivity or load capacity simply by printing a thicker or thinner elastic member. Key achievements include:

  • Cost Efficiency: A fully functional 6-axis sensor can be built for less than $20.
  • Accuracy: Most designs demonstrated errors well below 2% of the full-scale range.
  • Immunity to Noise: Unlike strain gauges, optical sensors are relatively immune to electrical noise from robot motors.
Detail of a 3D-printed sensor assembly, showcasing the integration of electronics and mechanical components.

The authors demonstrated several practical applications, including a 1-DOF humanoid foot sensor for center-of-pressure estimation, a 6-axis "bottle" sensor for studying pouring tasks, and an instrumented screwdriver for tracking human demonstrations.

Limitations and Future Outlook

While the software compensates for many mechanical shortcomings, 3D-printed plastics like PLA and ABS remain sensitive to temperature and exhibit "creep" under constant load. Additionally, the sequential A/D conversion in standard Arduino microcontrollers can limit sampling rates for high-sensor-count designs.

Future work points toward using digital I2C proximity sensors to further improve noise immunity and exploring hybrid manufacturing techniques where sensors and wiring are embedded directly during the 3D printing process.

Source Code: The design libraries, firmware, and ROS drivers are available at github.com/TAMS-Group/tams_printed_ft.

Summary updated

BibTeX
@article{hendrich2020d,
  author = {Norman Hendrich and Florens Wasserfall and Jianwei Zhang},
  title = {{3D Printed Low-Cost Force-Torque Sensors}},
  journal = {IEEE Access},
  year = {2020},
  doi = {10.1109/access.2020.3007565},
}
Figure