The study presents a queuing network model for optimizing zoning strategies in human–robot collaborative picking, but how does it account for real-world variability in warehouse layouts? Were any sensitivity analyses performed to assess the model’s robustness across different aisle configurations or varying robot-to-picker ratios?

As I was going through you paper, a few practical thoughts popped into my head that I figured I’d share. They’re mostly about how the theory might meet the messy reality of a warehouse floor.
1- The way you’ve modeled NZ, it seems like a picker could be instantly dispatched anywhere. But in a real warehouse, if my last pick was in the far right corner and the next robot is waiting all the way on the left, that’s a long walk! It feels like your NZ model might be a bit optimistic by not fully accounting for that “where is the picker actually standing?” coordination delay. Could that be making NZ look slightly better than it actually is?
2- You assume robots can easily overtake and never get congested. I totally get why you’d make that assumption to keep the model clean! But I’m picturing 40+ robots all trying to line up for the same popular zone or all converging on the depot at once. In practice, wouldn’t that cause some traffic and slow things down? I’m curious if this could affect the results, especially in those high-robot scenarios where PZ shines.
3- Your cost model is great for the “if everything is already running” case. But switching to a zoning strategy feels like it would need some upfront investment: new software logic, training for the pickers to stay in their zones, maybe even physical changes. I wonder how including those one-time setup costs would change the calculation on whether zoning is worth it.
4- Your conclusion that dynamic switching doesn’t help much is really interesting. My first thought was “but being adaptive has to be better!” But then I imagined being a picker and having my strategy change every 10 minutes. It sounds confusing and disruptive. Was that the kind of “deterioration” you found? I’d love a tiny bit more color on what “deteriorate” actually looked like in the model.