Large fleet AMR Simulation

[rhramac]

[ FlexSim 23.1.2 ]

I am trying to build a simulation model with 1000+ AMR drives in it. I am looking for some guidance on which path to go - A star or AGV network? A star implementation being slightly easier, I tried it out but would run into deadlocks. I was able to alleviate it a bit with preferred paths but completely successful. Are there any best practices or feature I could try to spread out traffic even if it is not the shortest path. Is it better to build an AGV path network and spread the traffic flow instead of using A star navigator?

Attaching a sample/test model

Test_Model_LS_AMR.fsm

[jason_lightfoot_adsk]

The model is running slowly due to the distance field being set to 'dynamic' - you could uncheck that and it will run faster.

Then for the deadlock it just looks like too many AMRs are at the pickup locations so you could try to create some zones to limit the numbers entering.

[rhramac]

1. Understood on the dynamic value on list.

2. Could you explain the zone to limit the number of AMRs? IS it similar to control Area? But my understanding is that control area works only with AGV network and not A star navigation?

[jason_lightfoot_adsk]

I was thinking in the process flow, once you know the destination, tell the amr to go to a buffer area and then try to enter a process flow zone before issuing the command to go to the final pickup/dropoff location. Have the zone(s) restrict the number of tokens by setting capicities.

[rhramac]

Restricting AMRs near pickup location did not help a lot as I am seeing deadlocks at different locations as well. From the heat map and just visually observing the traffic flow, I feel like I may be need to spread out traffic bit more evenly at cost of taking longer routes between source and destination. Refer image below- Red is probably shortest path, but I could divert some traffic to blue or Orange path to spread traffic out.

One thought is I would have some nodes and I would have to manage traffic by specifically routing AMRs through these nodes to get to the destination. I wanted to check in the forum if I am over thinking this or if there are any suggestions for me to try out?

[logan_gold · Accepted solution]

Hey Rahul, I did want to pass on some information about using an AGV network. While it would take more effort than implementing an A* network, using an AGV network does have these benefits:

1. With so many AMRs, using an AGV network will cause the model to run much faster than when using A*. Specifically, calculating routes should be much faster with an AGV network compared to a similar A* network.

2. Since you have to lay out the network beforehand, you will have more control with an AGV network. This may help with the deadlock issues you are having while using A*.

While building an AGV network, you do need to make sure there is enough room for all AMRs to be on the network at any given time. The default options for AGV objects is to have one AGV/AMR per control point/area, so you would need at least as many control points as AMRs. Too few control points could still cause deadlocks. Or you can also set up multiple allocations for some key control points, depending on how you want traffic to flow.

If you do want to pursue an A* network, you should be able to implement something like you've described in one of your comments, by routing the AMRs through some sort of network using nodes/dummy objects. I've not tested anything like this in your model, but it should work. If anybody else in the community has tried something similar, please feel free to weigh in.

I also wanted to mention that there is a feature coming in a future release of the software that will allow using Control Areas with A*, similar to how they are used with AGV currently. This may help the deadlocks in your model, but I'd be curious to see how it stacks against your modified model where you have already tried restricting AMRs near pickup locations. It may still cause deadlocks in other locations like you've already seen. However, since the feature is still in development, we won't be able to test it in this case until one of the FlexSim 2024 releases most likely.

[Jeanette_Fullmer]

Hi @Rahul R, was Logan Gold's answer helpful? If so, please click the "Accept" button at the bottom of their answer. Or if you still have questions, add a comment and we'll continue the conversation.

If we haven't heard back from you within 3 business days we'll auto-accept an answer, but you can always comment back to reopen your question.

[rhramac]

Thanks for the clarification Logan. I will try out both and see what works.