HELENE: Six-Axis Accessible Open-Source 3D-Printed Robotic Arm for Research and Education

Abstract

Robotic arms are used in a wide range of industrial and medical applications. However, for research and education, users often face a trade-off between costly commercial solutions with no adaptability and open-source alternatives that lack usability and functionality. In education, this problem is exacerbated by the prohibitive cost of commercial systems or simplifications that distort learning. Thus, we present HELENE, an open-source robot with six degrees of freedom, closed-loop position control, and robot operating system (ROS) integration. The modular design of the robot, printed on a commercial 3D printer, and its integrated custom electronics allow for easy customization for research purposes. The joints are driven by standard stepper motors with closed-loop position control using absolute encoders. The ROS integration guarantees widespread control options and integration into existing environments. Our prototype, tested in accordance with ISO 9283, has a small positional accuracy error of 8.4 mm and a repeatability error of only 0.87 mm with a load capacity of 500 g at a reach of 432 mm. Ten prototypes were built and used in various research and education applications, demonstrating the versatile applicability of this open-source robot, closing the gap between reliable commercial systems and flexible open-source solutions.

1. Introduction

Robotic manipulation has become increasingly important in automation [1,2]. Industrial applications have led to the development of many optimized robots for specific functions, environments, and technical requirements [3,4]. The diversity of robotic systems available meets a wide range of requirements and specifications, from handling heavy loads to achieving high precision. Some meet medical safety standards; others are certified to share their workspace with humans [5,6]. In educational settings, however, options are often limited to simplified models with reduced functionality [7,8] or expensive systems that limit accessibility [9,10]. Research applications are further hampered by the proprietary nature of many systems. While many commercially available robots support open software frameworks, such as the robot operating system (ROS) [11], the adaptability to individual requirements is often limited by the hardware [9,10]. Closed-source hardware and electronics prevent researchers from adding unsupported or custom sensors. One potential scenario is the development of new sensors for robotics, where both the hardware and software undergo modifications [12]. Several open-source projects aim to address these limitations, but mainly focus on cutting costs for private users and small businesses, at the expense of functionality [13,14,15,16]. Some researchers therefore developed new robots for their specific applications [17]. A feature that is often omitted due to these cost savings is closed-loop position control, which requires angular position sensors in every joint [18]. Relative encoders or even binary end-stops require calibration movements at each startup sequence. Some projects use closed-loop stepper motors, but do not use the data in path planning. Thus, disturbances are compensated locally without adapting the path. Other robots are developed specifically for research applications in the field of collaborative robotics and artificial intelligence, focusing on humanoid appearance and features such as serial elastic drives [19,20]. The electronics as part of the hardware of the listed open-source projects are often built around repurposed 3D printer control electronics because of their commercial availability but lead to limited expandability. Additionally, reliance on external electronics hinders future expansions [14]. Every extension to the robot requires rewiring, often redesigning the majority of the system. The non-commercial components such as housing parts are fabricated using traditional methods such as milling and turning metal, while some incorporate rapid prototyping techniques such as laser cutting acrylic sheets or 3D printing thermoplastics. Three-dimensional printing improves accessibility due to the affordability of commercially available machines and the ability to manufacture parts with one single piece of equipment. The ability to produce complex parts from a diverse, growing collection of materials has revolutionized low-cost manufacturing. For instance, this technology opens up completely new avenues in regenerative medicine, automotive manufacturing, and urban development, up to the printing of entire bridges [21,22,23]. This becomes apparent by these projects’ usually larger user communities, measured by the contributions and activity, e.g., on GitHub. However, load-bearing capacity and overall precision of printed parts are often reduced due to the properties of standard printable thermoplastics. We are not aware of open-source projects including metrological characterizations of the robots. However, this is an important feature, particularly in research applications, as it allows for the identification of potential limitations. In the field of low-cost robotics, many new systems are developed every year. While many of them certainly have comparable characteristics to the robot presented here, the data situation is insufficient for scientific work, as, for example, the measurement setup is not presented publicly.

In this work, we present our hands-on engineering learning and experimentation node (HELENE), a novel open-source robot arm design that can be constructed exclusively using a 3D printer and standard parts while supporting ROS (Figure 1). The project also includes modular open-source electronics, which are designed in particular for future expansion and customization. This paper is structured as follows. All data and instructions are available in the Supplementary Information. In Section 2, we present the system design, detailing the mechanical design, control electronics, and software components, followed by build instructions and operating instructions in Section 3 and Section 4. Section 5 outlines the measurement setup for a metrological evaluation of the capabilities and we discuss the results and implications of our work.

2. Design

The robot is designed for low-cost, robust, and simple use in various research and teaching tasks. Six degrees of freedom (DOFs) allow for full control of the position and orientation of the end-effector. All electronics are integrated into the robot, except for the 24 V direct current power supply. The use of an external power supply unit increases safety, as it is not necessary to work with mains voltage during installation or use. The overall size of the robot is 577 mm, and is, thus, similar to the commercially available models for teaching [10,15] and can be operated on a desk. Robots with a larger reach require larger actuators or transmissions, which in turn increase the dead weight, and, thus, the load on the printed parts used. All non-standard components are designed for 3D printing, requiring only one entry-level machine and no additional tools or machines. All remaining standard parts are selected for their wide availability.

Where possible, the same standardized bearings are used in the robot to keep the list of unique parts small. Absolute encoders on each joint reveal the position after startup and allow for closed-loop motor control. With knowledge of the absolute position at all times, no calibration is required at start-up or after a collision, making the system more robust for changing applications and users. The robot is controlled using ROS, thus integrated into a well-documented ecosystem for which a wide range of hardware and software exist that can be used together. In the following subsections, we provide a detailed description of the mechanical design, followed by an overview of the modular electronics board, and conclude with the software.

2.1. Mechanical Design

The serial robot consists of six rotational joints, numbered consecutively, starting with the base. Joints are interconnected by rigid links with only orthogonal joint axes. Offsets and relative joint positions are described using the standardized Denavit–Hartenberg convention, with four transformation parameters [24] (

Table 1. Denavit–Hartenberg parameters.

Joint	${\mathit{\theta }}_{\mathit{i}}$	${\mathit{\alpha }}_{\mathit{i}}$	${\mathit{r}}_{\mathit{i}}$	${\mathit{d}}_{\mathit{i}}$
1	${\theta }_1$	$-{90}^{\circ }$	0 mm	145 mm
2	$-{90}^{\circ }+{\theta }_2$	$0^{\circ }$	200 mm	0 mm
3	${\theta }_3$	$-{90}^{\circ }$	0 mm	0 mm
4	${\theta }_4$	${90}^{\circ }$	0 mm	177 mm
5	${\theta }_5$	$-{90}^{\circ }$	0 mm	0 mm
6	${\theta }_6$	$0^{\circ }$	0 mm	55 mm

). ${\theta }_1$ is the angle between the x-axes of two consecutive joints, measured around the actuated z-axis, and is therefore not constant. ${\alpha }_i$ is the angle between the z-axes measured around the x-axis, representing the rotation of one joint relative to the previous. $r_i$ is the distance between the z-axes along the x-axis. $d_i$ describes the offset along the z-axis from one joint to the next.

The size of the printed parts is restricted by the build volume of conventional printers. Subdividing parts necessitates additional connectors, and, thus, is avoided, since they introduce potential failure points. Threaded inserts are used to provide robust anchor points for bolts, ensuring future repairability. Standard bolts in sizes M3, M4, and M5 are used. Using multiple bolts helps reduce and evenly distribute tension on the plastic. When transmitting torque, the geometry is designed for a rigid claw coupling, ensuring the force is transmitted along the contact normal.

All printed parts are designed to be printed with Polylactic acid (PLA) for its good printability and strength. Every part is designed for a specific orientation in the printing process to reduce overhangs and prevent delamination of loaded parts. The build volume of the printer is required to be larger than 210 × 210 × 180 mm. For best results, we suggest a 0.4 mm nozzle at 0.2 mm layer height for maximum strength with five outer perimeters and 70% infill on a commercially available 3D printer (Original Prusa i3 Mk3s, Prusa3D, Prague, Czech Republic) [25]. This combination achieves high strength at reasonable printing times and filament usage.

Due to the serial structure of the robot, each joint has to move the weight of all following joints, thus causing different requirements for the various drives. Stepper motors are widely available in various sizes and allow for precise position control. However, maximum output torque correlates with dead weight, demands on the electronics, and the required installation space. The use of transmissions increases the design complexity and risks adding backlash in the drive train but provides a solution to keep motor sizes small. As a compromise, different NEMA 17 motors are used in all joints, with additional gearboxes and belt drives where required, except for the base joint where the motor is stationary and does not add to the moving mass of the robot. Here, a NEMA 23 motor (23HS30-3004S, StepperOnline, Nanjing, China) was used. While printed gearboxes can achieve high transmission rates, the wear of the plastic parts under load leads to greater backlash over time. Belt drives spread the load over a larger contact area than gears, but require tensioning and a large design space for large ratios. We therefore used stepper motors with premounted planetary gearboxes (17HS15-1684S-HG20/50, StepperOnline, China) combined with single-stage belt drives in joints 2 and 3. Only belt transmissions were used with stepper motors with NEMA 17 motors (17HE19-2104S/-2004S, StepperOnline, China) in joints 4 and 5; joint 6 requires no ratio at all. The combination allows for low backlash and simultaneously parallel offset of the motor from the drive axle with selectable transmission ratios (

Table 2. Gear transmissions.

Joint	1	2	3	4	5	6
Planetary		1:50	1:25
Belt	1:3.5	30:50	30:52	20:95	20:60
Combined	1:3.5	1:83.3	1:43.3	1:4.75	1:3	1:1

). Additionally, the belted transmissions are used to offset the heavy motors towards the previous joint (Figure 2).

Timing belts with 2 mm and 3 mm pitch are used depending on the transmitted force. Standard aluminum pulleys are used wherever feasible; special sizes are 3D-printed. Small ball bearings are used as tensioner rolls with set screws. On joint 3, the timing belt is additionally guided around a right angle, allowing the motor and the planetary gearbox to rotate and fit inside the body, keeping a slim profile and reducing possible collisions.

All joints, except for the sixth, are supported by two ball bearings, each in a fixed–loose configuration. At joint 6, the end effector is mounted directly on the motor axle, utilizing its internal bearings. Deep Groove Ball Bearings 6009 were used in joint one, and 6806 were used in the second, third, fourth, and fifth joints. All bearings are oversized relative to the expected loads, which helps to reduce stress due to the larger contact surfaces with the printed parts. The large inner diameters also facilitate the use of hollow printed axles with internal cable routing. The loose bearing is press-fitted into its corresponding hole during assembly and is designed to handle only radial loads. In contrast, the fixed bearing determines the axial position and bears all axial loads. Holes intended for the installation of bearings in printed parts are designed for an interference fit. These parts are also printed axially in the direction of the z-axis for the best roundness, guaranteeing an equal stress distribution. The exact tolerance is highly dependent on the printer’s accuracy and nozzle wear.

All wires are routed through the robot to protect them from collisions. Whenever possible, cables are fed through one end face of an axle, with the other available for the angle measurement. The cable is fed eccentrically at the other joints with designated buffer zones to increase the bending radius. Mechanical rotation restrictions on all joints prevent cable damage.

2.2. Electronics System Design

The electronics required are the same for all joints due to the standardized use of stepper motors. One printed circuit board (PCB) per joint is used to reduce the amount of wiring required in all moving joints, allowing for easy replacement, and extension of the robot. Each PCB is designed around a microcontroller (ESP32, Espressif Systems, Shanghai, China), a stepper motor driver (TMC5160, TRINAMIC Motion Control, Hamburg, Germany), an encoder (AMS5048A, ams AG, Premstätten, Austria), a Controller Area Network (CAN) interface, and an addressable RGB LED (Figure 3). The ESP32 was especially chosen for its wide availability, CAN-bus support, and second core, enabling parallel processing of various tasks. Internal communication is implemented using the Serial Peripheral Interface (SPI). The dimensions of the PCB are adapted to a NEMA17 stepper motor to directly mount it on the motor’s rear end, with the centered encoder chip on the PCB above the center of the rotating axis.

Measuring the absolute position of each joint with the magnetic encoder requires a diametrically polarized magnet at the end face of each axle. On joint six, the magnet sits directly on the rear end of the motor’s drive axle; on all other joints, the magnet is mounted to the printed axle. In joint one, a smaller sensor breakout PCB is used to place the encoder above the magnet. Positions are measured with 14 bit and provide a clear position due to the axle’s mechanical limitations to ≤360°. The accuracy of the angle measurement is therefore better than the motor clearance specified by the manufacturer.

The stepper motor driver was chosen for its integrated SPI interface, external MOSFETs, and the ability to directly command target velocities. The ESP32 directly commands the target velocities to the motor driver in order to streamline the control process and reduce latency. This approach eliminates the need for the ESP32 to perform additional calculations, such as velocity ramps, allowing the system to run faster and with lower processing overhead. A series of integrated control methods smooth the movement and reduce vibrations that are typically caused by square wave control. In contrast to applications such as in 3D printers where individual motors are only energized when the position changes, in robot arms, torques usually have to be applied continuously in order to maintain the current position. The resulting permanent load on the MOSFETs generates heat, which can be better dissipated by using external MOSFETs. The ability to control higher outputs also enables the use of larger motors, such as the NEMA23 used in the first joint, which could only be used in comparable models with additional performance-limiting measures.

The board is supplied by 24 V, which is required for the stepper motors and regulated down to 3.3 V for the logic devices, such as the ESP32 using a DC voltage regulator. With two additional wires necessary to implement CAN (CAN high and low), four wires are sufficient for the interconnections. Connectors for inputs and outputs are provided on the board to simplify the setup. Address select solder pads allow running the same code on each joint and identifying each board physically. The board used to drive joint 1 is the only stationary joint and therefore used to connect a computer running the controllers via the Universal Serial Bus (USB).

2.3. Software and Control System

Planning and control are conducted by a computer running Ubuntu 20.04 LTS with the ROS framework. The first joint’s control board implements ROS serial as an interface between ROS and the embedded firmware on the robot. This first board reads joint positions sent via the CAN bus from all joints and publishes them bundled in one custom message to the joint_position ROS topic. In the other direction, target velocities calculated by the controllers running on the ROS core are forwarded to the corresponding joints via the CAN bus. Every other board is only interfacing with the CAN bus. Here, incoming commands are redirected via SPI to the motor driver and, if used, to the LED or connected extensions such as an end-effector at the interfaces of the ESP32 microcontroller. Additional ROS topics, such as a gripper opening angle, need to be interfaced with the CAN bus on board one to be available on other boards.

The robot is controlled on the computer by the standard ros_control components. A custom hardware interface connects the robot to the joint_velocity_controlleer, implementing PID controllers for each joint. The robot’s geometry and kinematics are described using the unified robot description format (URDF) model, including a description of the kinematic tree, joint ranges, and mesh frames for collision detection. Using the standard controllers and hardware description allows controlling the robot with all integrated methods, such as the graphical interface RViz and robot simulator Gazebo. We further developed a Python interface to control some of the basic functions, such as point-to-point movements, allowing inexperienced users to program the robot with a few lines of code without having to forego the features of ROS. No deep ROS knowledge is needed to start programming HELENE.

Once the robot’s hardware has been assembled, an initial calibration sequence is required to determine the offset between encoder readings and joint positions. While the design limits the rotation to a known angle, the calibration process is used to find the maximum and minimum sensor readings for these joint limits and is performed with a firmware feature guiding the user, as described in the accompanying documents. By using absolute magnetic encoders, the position of the reference magnet only needs to be determined once and is then stored in the microcontroller’s permanent memory. The board of the joint to be calibrated is connected to a computer via USB. The user is guided through the calibration process via the serial interface to find the magnet’s orientation once press-fitted into the printed part.

The measured and calibrated encoder positions are fed back to the joint_velocity_controller to calculate target velocities. We applied the Ziegler–Nichols method for all joints independently to tune the PID values. The stiffness and overall behavior of the robot are significantly influenced by the assembly quality, tension of the belts, and printing tolerances. However, the differences between the ten prototypes built to evaluate repeatability are so small that it is not necessary to retune the controllers for each prototype. However, if an application requires it, retuning can be done quickly using a graphical interface included in ROS.

Due to the number of parts used in this project, the full bill of materials was added to the data repository together with visualized build instructions.

3. Build Instructions

The robot consists of numerous printed parts, as well as several standard components. A detailed set of build instructions for each assembly step, including wiring, is provided in the build instructions manual available in the repository. The manual is organized into sections, each dedicated to a specific assembly group, and lists all required parts for that section. Printed parts are referenced by their file names, and each part is accompanied by a rendered image for easy identification. Standard components such as bolts, washers, motors, and belts are listed both for each step and at the beginning of the document to ensure all necessary materials are gathered before starting the assembly process.

Instructions for compiling and running the software are provided in the repository’s README file. Due to the large number of parts involved, we will focus here on the general assembly steps and refer to the detailed instructions document for specifics such as individual bolt placement. The overall process, along with the approximate time required for each step, is outlined as follows:

• Printing Parts (5 days): This step requires a standard 3D printer (e.g., Original Prusa i3 Mk4, Prusa3D, Prague, Czech Republic). We recommend using PLA with a 0.4 mm nozzle, 0.2 mm layer height, four perimeter walls, and 70% infill with the standard PLA slicer profile. We recommend calibrating the extrusion beforehand to ensure that the standard parts, such as ball bearings, can be inserted properly in the intended positions without play. Depending on the printer used, support material is required for some parts and should be removed after printing. The repository contains a suggested division of parts into two colors, as well as the exported meshed files.
  Tools required: 3D printer and pliers.
• Preparing Electronics (1 h): This step is highly dependent on the manufacturer. We recommend ordering the boards pre-assembled using the provided production files due to the number and size of individual components. The code is uploaded to each board as described in the electronics’ README file.
  Tools required: Computer and micro USB cable.
• Assembly (8 h): This step involves connecting the printed parts and prepared electronics. Follow the detailed instructions in the repository. A vice or clamp is recommended for integrating the bearings with the required press-fit. Due to the integrated cable routing, it is important to follow the specified sequence described in the assembly guide, as certain areas will be inaccessible later without disassembling parts of the robot.
  Tools required: Pliers, hex keys for M3, M4, and M5 bolts, and vice or clamp.
• Software Setup and Calibration (2 h): The assembled robot features six integrated absolute encoders that require initial calibration. Each board is connected to the computer in sequence to perform the calibration steps. The process is initiated as described in the README, which guides the user through storing the home position for each individual axis. A video demonstrating the process is embedded in the README for reference. A computer running a standard ROS noetic installation is required. We have provided a comprehensive installation guide. For educational settings, we recommend using a portable system (e.g., a bootable USB stick) to facilitate usage across different computers without altering the main operating system.
  Tools required: Computer and micro USB cable.

4. Operating Instructions

The robot is controlled via ROS using the standard interfaces. The following steps outline the operation process: First, activate the 24 V power supply to power the robot. Second, connect the robot to the computer via USB. Third, start all ROS nodes using the commands listed in the README for the visual interface. A simulation can be used without requiring a physical robot, making it particularly beneficial for teaching. In this case, only step three is required.

Note that operating ROS is outside the scope of this work.

5. Validation

The robot is characterized in terms of positional repeatability and accuracy according to ISO 9283 [26]. We further tested the load-carrying capacity by adding weights on the outstretched arm until failure. Test poses are prescribed on the diagonal plane of the largest ISO cube possible in the robot’s workspace (Figure 4), aligned with the robot’s base coordinate system and with constant orientation (

Table 3. Test positions.

(mm)	Home	1	2	3	4	5
x	247	326.2	36.3	181.5	326.7	36.3
y	0	$-145.2$	$-145.2$	0	145.2	145.2
z	345	76.3	366.7	221.5	76.3	366.7

).

The same sequence was performed a total of 30 times, each consisting of three cycles through the prescribed positions at 100 %, 50 %, and 10 % of the maximum possible velocity. ROS was used to both control the robot and to trigger the measurements through the Transmission Control Protocol (TCP). Reference measurements of the positions were conducted using a coordinate measuring machine (LH87, WENZEL Group, Wiesthal, Germany) with a probe head (PH10, Renishaw, Wotton-under-Edge, UK) against a reference cube with an edge length of 15 mm attached to the end-effector (Figure 5). The robot was mounted directly on the test-machine bed. All measurements were conducted in a controlled environment at 20 ± 2 °C. The positional accuracy and repeatability were calculated from all cycles according to the procedure defined in [26].

Results and Discussion

The positional accuracy was measured at the five prescribed positions, with an average error of 8.4 mm (

Table 4. Positional accuracy.

(mm)	1	2	3	4	5	Average
x	$-1.615$	5.228	0.204	$-8.302$	6.085
y	9.178	$-3.643$	$-4.362$	$-0.038$	10.397
z	$-2.919$	2.646	$-2.133$	0.776	$-1.148$
total	9.765	6.900	4.860	8.338	12.101	8.393

). The discrepancy is greatest at positions one and five, where the arm is in a widely extended position. In horizontal stretched positions, the robot’s weight places a significant load on joints two and three. Additionally, small angular deviations in the calibration of the first joints are propagated with the lever of the outstretched arm.

The repeatability of pose is approximately ten times better than the positional accuracy with an average deviation of 0.87 mm, supporting the assumption that calibration has a significant influence (

Table 5. Repeatability.

(mm)	Home	1	2	3	4	5	Average
x	0.14	0.269	0.256	0.089	0.688	0.754
y	0.173	0.155	0.116	0.111	0.837	2.556
z	0.255	0.131	0.111	0.112	0.69	0.813
total	0.338	0.337	0.302	0.181	1.285	2.786	0.872

). The encoders’ resolution of 0.05° results allows for a maximum precision of ±0.38 mm at the outstretched position, resembling the outcome for test position five. In the horizontal poses, the repeatability is better despite poorer absolute accuracy. Here, gravity influences the final position, yet preloads the joints in such a way that the backlash has less effect than in vertical positions. The opposite effect is seen in poses where the center of gravity aligns with a joint’s center, such as in outstretched vertical positions. Here, the specific joint is unloaded and the backlash error is fully manifested. Deviations are most likely caused by the free calibration and remaining flexure of the joints. Other replicates of the robot can have different properties due to their calibration and the quality of the assembly, e.g., the tension of the belts. However, our ten exemplary robots are built using different printers and show comparable results regarding the load capacity and repeatability. Variations between robots are influenced by several factors, primarily including the individuals involved in the setup, the specific printer used, and the calibration process. If a highly precise robot is needed for a particular application, the properties demonstrated here may only be applicable to a limited extent. In such cases, a custom-built system may not be suitable. Comparability of numerical values with other systems is difficult to achieve, as the open-source systems in particular have mostly been characterized according to their own tests. In most cases, only the repeatability is determined with a reduced measurement setup, i.e., the deviation when repeatedly approaching a single point, which is also not tested at multiple positions in the workspace as defined by [26]. In most educational applications, accuracy plays a subordinate role and the focus is primarily on robustness, cost, and functionality. In addition to the comparability of systems, performance metrics are particularly important in research tasks, where the error must be quantified. The metrics allow users to make estimates in highly individualized scenarios—something that would not be possible if the system is not sufficiently characterized and quantify the error.

HELENE is able to manipulate a payload of 0.5 kg in all positions and orientations, comparable to other commercial and open-source robots [10,13]. The primary limiting factor was found to be the sixth joint, which experienced step loss under heavy loads. However, no permanent damage occurred during the overload tests. In some applications, payloads up to 1.5 kg are feasible with a limited working space by restricting the sixth joint’s orientation. While this payload capacity might be increased by using higher transmission ratios or more powerful motors, it is important to note that larger and heavier motors would directly contribute to the overall dead weight. This increases the load on the preceding joints and the printed components, affecting both performance and structural integrity. The load capacity is more than adequate for our applications and is also suitable for use cases such as described in [17].

6. Example Applications and Conclusions

We present an open-source, 3D-printed robot arm with six degrees of freedom and a maximum reach of 432 mm for research and education. The costs for the hardware amount to less than 1000 € in the year of 2024 per robot, which is a fraction of comparable commercial systems. Approximately 50% of the costs are for electronics and 50% for other hardware. Notably, the cost of the electronics decreases significantly when multiple robots are produced, especially through the automation of electronics production. Ten units are already in use and provide a reliable platform in many different use cases (Figure 6). In a laboratory exercise, all students can learn the fundamentals of robotics with hardware, where originally simulations were used to a large extent for cost reasons [27]. The robots have successfully endured $\sim$500 h of operation and usage by $\sim$50 students, demonstrating their reliability and robustness. A study was conducted that demonstrated the advantage of a hardware-based approach to teaching the fundamentals of robotics. Furthermore, the robot is used for various tasks in research projects, such as robot-assisted biopsy and palpation [28]. Here, the robot is used to guide a sensor-integrating needle. In contrast to comparable commercial products, the open hardware allows for easy adaptations. One exemplary use case is the repetitive positioning of measurement equipment to investigate the spectrum of visible light. Here, all parts are printed in matte black to reduce measurement interference. The modular electronics are extended by one additional module used to control a seventh motor in the setup.

In future work, the robot design itself will be improved by incorporating more ROS functions, such as force control and ROS2 compatibility. Recently, multi-material 3D printers have become more accessible as well as printers able to print composite materials. We will use this to integrate printed sensors and reduce cabling efforts through conductive filaments. We will also evaluate how the remaining flexure is reducible using lightweight nanocomposite materials [29,30]. Therefore, we already developed a versatile sensor for integration in printed parts (Figure 7). The expandable electronics are beneficial, adding the required readout circuit without rewiring the robot. Further integrations of sensors in the robot’s joints are possible due to the printed axles. Additionally, we will examine the system’s potential applications in further educational and research contexts.