How to initialize particles randomly to get the first population in Particle Swarm Optimization?

I'm applying PSO to an Inventory Routing Problem, and the solution consists of a list of shifts to delivers customers. A shift is defined by: --The driver's index --The index of the vehicle used --The start time --A list of operations

--> Each operation is defined by the index of the customer, the arrival time at this customer and the quantity delivered.

So to initialize the swarm, we must set random values to each variable in the range of its possible values ?

And I know the initial population must contain feasible solutions, but if we generate values randomly for each variable, it's not gonna be the case..

Please any help is appreciated!