question

Here's a demo model I made:

RouteOptimizationDemo.fsm

The key to solving a routing problem is to use a Sequence parameter. A Sequence parameter is an array of values that represent a sequence. Your example has 4 destination cities, so we might consider a sequence with four values: [1, 2, 3, 4]. Other possible sequence values might include [3, 1, 2, 4] or [4, 3, 2, 1]. So a sequence parameter of length n contains all values [1, n] exactly once, but they can be in any order.

So now we just have to map the sequence array to a route for a truck. The way I did it was to add a Global Table with each of the destinations, and how many tons were to go to each destination. Then, in the Scheduled Source in Process Flow, I created one token per row in the table. This is where I used the sequence parameter. The first token uses the first element in the sequence parameter as a row number. Then it reads the name and demand for that city. The second token uses the second element in the sequence parameter, and so on. So the tokens read the table in the order specified by the sequence parameter.

So now, I know the order that I want to visit the cities in. But how to make sure each truck only carries 10 tons? I did that by adding a Zone. The Zone uses a Subset, and only allows 10 tons of demand into the zone at a time. The Zone also enforces the queue order, so that "smaller" tokens can't sneak past "larger" ones (in terms of demand).

Then I use a Batch activity. The batch collects as many tokens as it can (in zero time), and then release the batch. When the batch is released, I used an Event Triggered Source to create a token for the batch as a whole. The batch token reads the list of tokens in that batch, and builds a task sequence to visit each destination. But also, since the batch is release, all those tokens leave the Zone, allowing the next few tokens in.

I also made two performance measures: the time it takes to make all deliveries, and the total distance the trucks travelled.

With the sequence parameter and these performance measures, I added an optimization job. It quickly tries all combinations of routes.

Note: this seems like a good place to use GIS, to get better time estimates between destinations. I didn't do that, but the same approach would work. It would just use the GIS navigator instead of the AStar navigator. Also note that the position and size of barriers affects the final answer. And finally, note that I made a whole bunch of assumptions about how to go from a sequence to routes for multiple trucks.