Designing an Omnidirectional Mobile Robot: Confusion Arising from KUKA Omnirob Model

[assaduz22]

Hello,
I'm quite new to CoppeliaSim and am in the process of designing my own omnidirectional mobile robot within it. To guide me, I've been looking at the KUKA Omnirob model as an example. However, I've found myself perplexed by the terms "frontJoint" and "frontLink." My questions are:
a) Why does the model include this additional joint? I understand its role in connecting the "frontLink" to the Omnirob body, which leads me to my next question.
b) What is the purpose of the "frontLink"? Why aren't the front two joints directly attached to the Omnirob body?
What are the advantages of incorporating this extra joint and link?

[fferri]

Not sure about the kinematics of the real robot, but in that CoppeliaSim model that seems to implement the solid axle, i.e. a rigid axle with a perpendicular joint allowing for axle roll, which is generally used to ensure better contact with terrain (otherwise, with rigid body simulation engines, there would be typically only 3 out of 4 contact points).

[coppelia]

The reason is following:

A real-world mobile robot (or a real-world chassis, beam, etc.) always has some compliance. Additionally, a real-world vehicle often has some suspension or tires.

With a physics engine, the compliance does not exist, or is very very small and due to numerical imprecisions, etc. So a typical model of a 4-wheel vehicle will be hyperstatic on the floor, i.e. at a given time t, only 3 out of the 4 wheels will touch the floor. This results in strange behaviour (e.g. non-straight trajectory with all motors at the same speed, etc.). One solution would be to add some suspension to each of the wheels, but another much simpler solution would be to add that passive joint, which allows the two front wheels to act as a single contact point on the floor, a little bit like a mobile hanging from the ceiling.

Cheers

[assaduz22]

Thank you everyone for the clarification.

I have another question about the same model. I noticed that the robot's dynamic body consists of multiple rectangular boxes grouped together as a single object. Why is this done? Does grouping them in this way offer better stability than using a single rectangle, or is there another reason behind this approach?

[coppelia]

A random mesh never performs well when a physics engine simulates it in collision response. The reason is that in order to be able to compute precise collision response forces, a penetration depth needs to be calculated between the two colliding objects. With random shapes, there is no such penetration depth and so collision response calculations are slow and jittery/poor. To avoid this you have 2 options: model the underlying shape as a convex object, or as a convex compound. Also, a convex shape can be a random convex shape, or a primitive convex shape, such as spheres, cuboids, cylinders, etc. The latter usually perform better than the former.

Cheers

[assaduz22]

Thank you for the quick response.
I'm interested in learning more about Mecanum wheel design. I've already looked into other discussions about the design strategies for such wheels. However, I noticed that the dimensions of the inner sphere and the wheel sphere differ. For the KUKA youBot, the dimensions are "0.09m for the inner sphere" and "0.1m for the wheel sphere." For the KUKA Omnirob, the dimensions are "0.216m for the inner sphere" and "0.25m for the wheel sphere." My question is, how does one determine the size of the inner sphere? Can it work with any dimension smaller than the wheel sphere?

[coppelia]

Only the wheel sphere dimensions are relevant: the insideSpheres are not respondable (i.e. they do not react to collisions), and thus only their mass and inertia matters.

Cheers

[assaduz22]

Thank you for the clarification. Given this information, could you assist me in determining the mass of the spheres in my situation? If the mass of my mobile robot's wheel is 5kg, what should be the mass of the inner sphere and the wheel sphere? Additionally, the mass of my robot's body is 30kg. Based on this, how much should the front link weigh?

[coppelia]

I'd go with a mass for each of the spheres of 2.5 kg. Use the same mass-less inertia for both. If stability is not good with the used engine, increase their mass and/or inertia somewhat.

Cheers