July 6, 2017 - July 6, 2017
3:30 pm – 4:30 pm
S.W.F. Jansen MSc and P. Sun MSc
An adaptive large neighborhood search heuristic for the time-dependent profitable pickup and delivery problem with time windows
The time-dependent profitable pickup and delivery problem with time windows is the generalization to the case of pickup and delivery problem. In this problem, each request consists of a pickup location and a delivery location and a profits is collected from the visit to its pickup location. A limited amount of vehicles with capacity limit are available to serve the requests. The profit of a request can be collected by one vehicle at most. The objective is to determine a set of tours with departure times at a depot such that maximize the difference between the collected profits and the total routes duration cost. Time-dependent travel times are considered to capture road congestion. We propose an adaptive large neighborhood search (ALNS) algorithm for this problem. Results of extensive computational experiments show that the ALNS is highly effective in finding good quality solutions on the generated TDPPDTW instances with up 100 freight requests that reasonably represent real life situations.
Optimal planning of stochastic activities in a production network
The production of high value, customer specific products, such as aircrafts, consists of a large number of activities. Since each activity has precedence relations, we model the production network as a Directed Acyclic Graph (DAG), where each node represents an activity. Each node has a stochastic leadtime. We incur holding costs from the moment an activity is started, until the delivery to the customer. If the product is not finished at the due date, penalty costs are incurred. Our objective is to minimize the total expected costs, by determining the optimal planned start time for each activity. We show that for a specific set of nodes a Newsvendor fractile can be derived, which denotes the probability that under the optimal solution, a specific activity in the network causes the final product to be finished late. To determine the planned start times, simulation based optimization is used. We show that our method gives near optimal results in a short amount of time for networks of significant size.
Location: TU/e, Building 62, Paviljoen F9. Feel free to join if you are interested!
Eindhoven University of Technology