Welcome to our mini workshop today on robot learning, optimization and control.

So we have four speakers today.

We have Morad, Elisa, Nora and Andreas and each of the talks will be for 30 minutes and

we have this short break of 15 minutes after the first two talks.

Please let's try to stay on time so that we don't take too much time today and of course

we will have sufficient time for the questions and so on, but let's try to keep on time.

Okay, so Morad is from FRO who just finished his PhD thesis and then he will be speaking

about processing time optimization for military robot system using heuristic algorithm.

So please Morad, you have the floor.

Okay, thank you very much.

So good morning everyone and welcome to this mini workshop and as introduced my name is

Morad Moradi and I'm delighted to talk to you about the topic of my PhD which is processing

time optimization for multi-robot systems and I'm using heuristic algorithms.

And let me briefly go through the agenda.

So first I will introduce you my use case and the corresponding optimization problem.

Then I will show you my methodology for the collision detection for the motion planning

and I will also show you my optimization algorithm.

And after that I will present you with the results of my experimental evaluation.

And finally I will sum up the most important points again.

So let's start with the use case, the PVC sealing process.

This is a very important process for the corrosion resistance of the car body.

In this process we have multiple robots and these robots apply a sealing material called

PVC to several parts of the car body as seams and after the robots applied all the seams

to the car body the car is driven through an oven where the sealing material hardens

and then you can just paint over it and then the seams are hardly noticeable for the customers.

So this is a very complex process as you can see then in the next slide.

So basically you want to calculate a production plan in which all seams are processed within a

minimal time span and for this you have to master two optimization tasks.

First the scheduling where you want to efficiently allocate the seams to the available robots but

at the same time you also have to solve the sequence and the motion planning where you

want to find the optimal path for each robot to process these seams.

And the main goals for my thesis is shown here below.

So I have three main goals.

First of all I want to automatically generate robot programs.

Second of all I want to increase the efficiency of existing production facilities and third of all

I want to provide a key value with which I can estimate in a very early stage of a new car

model's development phase if this car model can be integrated in the existing production facility.

And now let's have a look how I model this process and first of all I have to model the robot cell.

For this I need the cell periphery with the conveyor system.

Then I need the robots including the linear axis and the tooling and I also need the car body.

Then I have to integrate the robot task data and for the BPC seams this includes the seam center

lines to know where the seams are positioned and also I need the spraying vector to know how the

robots should align the nozzles to apply the seams on the car body.

And for the collision detection I also need some envelope bodies and I have simple box colliders

for unimportant cell periphery which are far away from action for example the linear axis

and I have capsule colliders for dynamic objects for example the robots

because the capsules can get the shape of the rotational symmetrical robot arms very well

and also the capsule colliders can be updated very quickly.

So they are very suitable for dynamic objects and for complex shaped bodies for example the car body

I'm using a point cloud collider and this is now the model of the BPC seam I'm using