MDVRPDRL_Data

The 2D coordinates of depots and customer nodes are uniformly and randomly sampled within the square region of [0,1]×[0,1].
The cargo demand of a single customer is an integer between 1 and 9, which is generated by uniform random sampling with no additional constraints.
The vehicle capacity is dynamically matched with the customer size, with 20-customer instances corresponding to a capacity of 60, 50-customer instances corresponding to 150, and 100-customer instances corresponding to 300. Only two configurations of depots are supported, which are 2 depots or 3 depots, and only two configurations of vehicles per depot are supported, which are 2 vehicles or 3 vehicles.
The time window range is divided by customer size, with [0,10] for 20-customer instances, [0,20] for 50-customer instances, and [0,40] for 100-customer instances. The time window of a single customer includes a ready time and a due time, with the ready time uniformly sampled within the corresponding range. The duration of the time window is 1.0 to 5.0 units, and the due time does not exceed the upper limit of the time window for the corresponding customer size to ensure that the ready time is less than the due time.
Both the early arrival penalty factor and the late arrival penalty factor are uniformly and randomly generated within the interval [0,1], which are used to quantify the violation cost of soft time window constraints.
This dataset does not include the service time of customer nodes, and all distribution actions have no additional time consumption constraints.