Originally posted by: AndyHam

Suppose a vehicle picks up and delivers jobs among machines in a factory line. One vehicle is currently located at machine 3 after delivering j1. Now, the next job is j2. So, the vehicle should first travel to m1 and then travel to m2 as shown in Fig. 1a. However, my CP code only represents Fig. 1b scenario, which is an incorrect representation. 
We can divide each job into two sub jobs to correctly model, but I think it is not a good approach. Please give me some idea, as IBM always does.
============================================   
j in Jobs;
m in Machines;
v in Vehicles;
tuple t_Job {
  key  int job; 
  int        from_machine;
  int        to_machine;
};
{t_Job} Jobs = ...;
tuple t_m2m_Travel {
key int   m1;
key int   m2;
int ttime;
};
{t_m2m_Travel } m2m_Travel;
dvar interval itvJobs[j in Jobs] size t_m2m_Travel[j.fromMachine,j.toMachine];
Sequence seqVehicles[v] in all(j in Jobs) types j.toMachine;
forall (v in Vehicles) noOverlap(seqVehicles[v],m2m_Travel);
============================================   

Originally posted by: PhilippeLaborie

As you guessed, you need additional interval variables for the jobs in your model. In fact you have two types of resources: the machines and the vehicles. And for a given job, you have 3 time intervals of interest for a job executing on a machine M: the delivery by the robot of the job at machine M, the processing of the job and the pickup of the job to the next machine. The moves of the vehicle would be modeled by transition times on a noOverlap constraint on the pickup/deliveries. On of course, you have the noOverlap on the processing intervals on the machines.

Philippe

Originally posted by: AndyHam

Thanks for the advice. Let me start from here. I am working on "Scheduling of Drone-Based Material Transfer System"