Modular testbed for evaluating soft robot's controllers

Alongside the advancements in soft robot technology, there is also a need for a standardized hardware platform that can enable benchmarking of various control methods developed for soft-bodied robots. This modular testbed contributes to the state-of-the-art by allowing the researchers to evaluate their control methods on standardized and open-source platform that that features exemplary tendon-driven soft-bodied robots with integrated closed-loop tip position and tendon position and force control.

Name of Principal Investigator: Maja Trumić
Position / institutional role: Research Associate
Info Email: maja.trumic@etf.rs
ORCID persistent identifier (PID): 0000-0002-2945-8010
Name of Host Organization  University of Belgrade – School of Electrical Engineering (ETF) 
Department or Lab   Department of Signals and Systems, ETF Robotics lab 
Name of Building  Palace of Science 
Physical Address   Kralja Milana 11, 11000 Beograd, Serbia 
Website Links  https://robot.etf.bg.ac.rs 
Institutional contact name  Prof. Kosta Jovanović
Institutional contact email  kostaj@etf.rs
Non-academic stakeholders
Startups; SMEs
Academic stakeholders
Undergraduate students; PhD students; MSc students; Researchers
Other types of stakeholders
Application case: Short description:
Benchmarking soft-robot control methods The TestBed is intended as a reproducible hardware platform for comparing control methods for tendon-driven soft continuum robots, including model-based and learning-based approaches.
Education and student projects Suitable for teaching mechatronics, embedded systems, soft robotics, CAD, PCB design, sensor calibration, control engineering and experimental data analysis.
SME/startup support and rapid prototyping The TestBed can support early proof-of-concept work for compliant manipulators, soft grippers, safe human-robot interaction devices and low-cost robotic actuation modules.

list of hardware components with their brief descriptions:

Modular soft-robot TestBed frame — 20×20 mm aluminium extrusion frame with stationary base, movable frame, base plate, hinges/locking profiles, T-nuts and M4 screws. The central part can be adjusted from 0° to 90° to support tests with upright or highly elastic soft bodies.

Soft continuum robot bodies — Silicone-based continuum bodies for tendon-driven soft robotics experiments. Supports cylindrical, cuboid and other rod-like shapes; supports internal and external tendon routing. Mold Max 30 silicone, Shore hardness 30A, is used in the documented version.

Universal lower and upper plates — 3D-printed PLA plates used as the mechanical interface between the frame, silicone body and tendons. Plate design includes anchoring points, button-head screws, coned screws and PTFE tendon guides.

Plate holder and tendon-routing hardware — Pneumatic adapters and PTFE guide tubes used to guide tendons with low friction, protect tendons from wear and enable quick tendon replacement.

Modular casting molds — Mold systems for internally and externally routed tendons. Internal routing uses a thin-walled metal pipe with 50 mm inner diameter and 3D-printed holders; tendon channels are formed using 2 mm wires. External routing uses a reusable fully 3D-printed mold.

Custom tendon actuator modules — 3D-printable actuator units with actuator mount, motor mount, tendon spool, integrated load cell, motor driver electronics and instrumentation amplifier. Designed for standalone use or multi-actuator daisy-chained setups.

Brushed DC motors — Documented setup uses 12 V Pololu 19:1 geared brushed DC motors; electronics support 8–28 V operation and up to 5 A motor current, enabling 12 V and 24 V brushed DC motor variants. Only the motor-mount part needs redesign for alternative motors.

Custom actuator PCB — 4-layer PCB with Raspberry Pi Pico microcontroller, DRV8842 motor driver, INA826 instrumentation amplifier, I2C connectors, power connectors, motor connector and 5 A fuse.

Force sensing and calibration equipment — Load cell integrated into each actuator for direct tendon-force feedback; calibration stand with reference/calibrated load cell, tensioner and spring for mapping amplifier voltage to force.

IMU sensor — MPU6050 IMU mounted at the robot tip in documented experiments for measuring pitch/roll angle and evaluating tip-orientation tracking.

Power and communication equipment — USB connection for single-actuator control, I2C bus for multi-actuator daisy chaining, serial interface to a PC, and external DC power supply suitable for the selected motor voltage/current.

Other recommended safety hardware — Current-limited bench power supply, emergency stop or master power cutoff, physical tendon/rotating-spool guards, PPE for silicone handling, and clearly marked operating area.

list of software components with their brief descriptions:

Actuator firmware — C firmware running on Raspberry Pi Pico for low-level motor control, load-cell reading and PI force control.

Soft Robot Controller firmware — Raspberry Pi Pico controller acting as I2C master in multi-actuator setups, translating serial commands to actuator-level I2C commands and collecting actuator/IMU data.

G-code-inspired serial API — Command interface with commands such as stop all, enable/disable driver, change decay mode, change control mode, manual drive, and set force reference.

Python serial-control scripts — Python scripts for controlling the robot over serial, sending actuator references, logging test data and saving CSV files.

MATLAB and Python evaluation scripts — Scripts for benchmarking control performance using RMSE, maximum absolute error, rising time, settling time, overshoot, tendon-position error and tendon-force regulation error.

CAD and mechanical design tools — Autodesk Fusion 360 used for frame, actuator, plate and mold design; 3D-print slicer software required for producing PLA parts and molds.

PCB production files/tools — Gerber files for custom actuator PCB; PCB viewer/manufacturing workflow required for reproduction.

PC environment — Any PC operating system capable of USB serial communication and Python execution. Exact OS, Python version and library versions are not specified in the provided README/PDF and should be documented locally.

Recommended Python stack — pyserial, numpy, pandas, matplotlib or equivalent tools for serial communication, CSV handling and plotting. Exact dependencies should be confirmed from the repository scripts.