IRAS - Intuitive Robotic Arm System (DIY 6 DoF Robotic Arm)

WORK IN PROGRESS!!

IRAS is a 6-DoF robotic arm completely designed by myself with which I will compete at Jugend-Forscht, Germany's largest science and robotics competition, in 2026. This robot is my second robotic arm and my third large robotics project.

IRAS is my most advanced robot yet, as I set myself some fundamental objectives:

1. easy and intuitive to use and control
2. high precision and high rigidity
3. payload capacity of at least 10kg
4. large work envolope
5. complex wrist assembly

I started this project in March 2025 and will continue working on it until it is completely finished.

For more details and other robotics related projects please check out my website at:

jacobutermoehlen.com

And thank you to JLCCNC for machining the aluminium parts as well as to NABTESCO for providing me with strain wave gears.

I'm very gratefull to have some awesome sponsors - this project would not have been possible without their generous support:

1. Nabtesco, who kindly gifted me two harmonic drives for joint 2 and 3
2. JLCCNC, all of the CNCed aluminium parts were machined by them

(As of now, August 2025, not every part of the robot is finished in CAD, I concentrated on the first 3 joints as those are made of aluminium by JLCCNC)

The CAD design was one of the harder steps of this project, eventhough I had 1.5 years of experience with CAD, as I designed "smooth and organic" parts for the first time.

I set my self design goals I followed:

1. parts made of aluminium should be easily machinable for reduced costs
2. 3d printed parts need to hide parts of the robot, to make him sleek and good looking
3. minimize visible screws where possible
4. design with high rigidity in mind
5. the robot should be easy to disasseble at two points, to make transportation more comfortable

I started out by defining the overall dimensions of this robot including a work envelope. Then made my way from bottom to top. Joints 1 to 3 are all made of aluminium for extra strength and rigidity. The upper 3 joints will be mostly made from 3d-prints to keep weight down near the end, because every gram further up the arm will decrease its payload capacity.

As seen in the 2nd image, I took inspiration from other industrial robots and placed all stepper motors for joints 4 to 6 behind the 3rd joint. I did this to decrease load on the joints 2 and 3, and can therefore have move usable payload.

The mechanism, why this works is rather complex, I will go into details in one of the next chapters.

Eventhough I didn't count the hours spent designing the robot, I can assure that those were alot of them.

With the assembly seen above I can decouple all motors from the corresponding joint actuators (Joints 4 to 6).

On the right side is the actuator for joint 4. This actuator is a 3d printed strainwave gear with a hollow shaft, and an off-centre input. This is neede to pass through two shafts for joint 5 and 6.

The shafts for joint 5 and 6 are concentric (the inner one is for joint 6 and the outer one for joint 5), this is achieved by the coupling block on the left. There the upper and middle stepper motor outputs are are connected to their corresponding shafts. The middle stepper is directly connected, while the upper one used a belt.

Through two universal joints, the bottom stepper motor will be connected to the off-centre input of the joint 4 actuator.

I opted for an interference fit between the bearings and the machined parts to eliminate any play and create a strong connection. The holes in the bearing housing were on purpose 0.05mm undersized and I heated them to around 60°C in the ove, to expand the holes. The outer bearing races were put into the freezer.

Freezing the spindle (top left image) was a mistake, once exposed to air, condesation started to form and create and thick ice layer. I let the part warm up to room temperature and heated the inner bearing races to 85°C and assembly worked perfectly.

Recommendations: Freezing smaller parts is better than big chunky ones, they heat up quicker and leave near to no condensation when mated with warm parts.

I repeated the same assembling procedure for joints two and three.

Currently I joint three is belt drive. However I have to test how well this works and eventually skip the belt and make it directly driven, which wouldn't be that complicated to change afterwards.

Eventhough I use industrial grade harmonic drives for the first three joints, I dont use off the shelf ones, because I wanted to keep cost down and also have more design freedom.

In the past I have already designed gearboxes for my first robotic arm, though they were fully 3d-printed and lacked precision and strength. For the design of the 6th gearbox I focused on:

1. as little backlash as possible
2. high torque
3. torsional stiffness
4. high bearing loads
5. compactness

My first design (first image) was neither capable of high torque nor designed for high torsional stifness.

The area circled in blue is the major weakpoint of this iteration. The pins responsible for driving the output are only press fitted 5mm into the material and during test often bent the hole, especially as they had to be relatively thin and long.

(My test were conducted with plastic parts, though simulation have shown, that even in aluminium the hole would deform reasily under high load.)

This this design is therefore not suitable for high loads and good stiffness. The red area is wasted space, which I tried to eliminate in the 2nd iteration.

Promising seemed to be my 2nd version, eventhough its called the predecessor, the design is nowhere close to the first. The make the assembly more compact, I moved the mechanism between the two tapered roller bearings, this way the otherwise wasted space (like in V1) is beeing used effectively.

Also the output driving pins, can no longer twist as easily, as they are supported from the back and also screw into the output which decreases bending.

After I have finilized my design with more than 10 prototypes, I let JLCCNC machine the critical parts for me from aluminium 6061 and 7075. JLCCNC did a great job and through their help and precise machining the final gearbox has zero backlash and a high rigidity.