PVRP - possible to tacke with OptaPlanner?

[Lukas Himmelreiter]

My planning challenge focuses on creating set period schedules, say 12 weeks, and methodically scheduling deliveries of forecasted demand in order to optimize fleet capacity utilization in this period. I can forecast the demand pretty good, those numbers are given and are made available to the algorithm through „strategic assignments“. Unlike traditional VRP, which optimizes routing based on existing orders, PVRP uses expected demands and strategically allocates resources. It incorporates varying customer sizes, size of demand (say in boxes or containers per period), delivery frequencies, and preferences into a cohesive plan which ensures efficient fleet utilization while accommodating specific delivery patterns and customer preferences. Some may only have one delivery day per week, others may be delivered to daily, some may have patterns, such as "Monday Wednesday Friday", whereby the second days are flexible and could be moved to "Thursday Saturday", while "Monday" is fixed. This model must balance the logistical constraints of fleet capacity with the nuanced demand of customers, who may require deliveries on fixed or flexible schedules. The aim is to distribute deliveries in a way that maximizes service efficiency without exceeding or underutilizing fleet capabilities, taking into account the strategic importance of maintaining consistent service levels across varying customer segments. This scenario is a specific application of the Periodic Vehicle Routing Problem (PVRP), where the objective is to distribute deliveries over a set period based on predetermined i.e. exptected demand, ensuring optimal fleet utilization at any given point in time, or a signal to increase/decrease fleet size or at least the available vehicles at certain points in time. Once this strategic plan is calculated, customers will only be able to place orders for delivery on a day where they or their delivery area as been scheduled in the plan. On all other days, delivery is (generally) rejected. On an effective day, say some monday, the actual plan can be end-optimized by a classical VRP with constraints. It is crucial to differentiate between the strategic tour planning (PVRP) and daily routing optimization (Classical VRP). Initially, strategic tour planning, or PVRP, involves distributing forecasted demand into an efficient delivery schedule over a set period, considering customer preferences and fleet constraints. Subsequently, for each day, a classical VRP optimizes the actual deliveries based on the day's specific order pool. In short, the Strategic PVRP distributes expected customer visits and amount/weight/volume of goods to certain days, while the classical VRP takes the days concrete orders and creates routes for vehicles to drive along the customers.

Tools like OptaPlanner or Google OR Tools are obviously well-suited for solving daily VRP challenges. However, the capability of these tools to model and solve the PVRP as described, integrating strategic planning with daily operational adjustments, is something I am looking into. Would you reckon this problem could be modelled with OptaPlanner?

[Petr Jandal]

Hi Lukas,
I am working on my final thesis, maybe we can find some synergies.
Is it possible to send you direct message?
My email is janda...@gmail.com

Best regards.
Petr