10h30 - 10h55
A stochastic production-distribution problem with a choice of transportation lead time
In general, a factory forwards its production to various retailers spread over a large territory, and can thus call for various freighters. For example, a production facility could call upon air, road, rail or river transport companies. These different transportation modes have different costs but also different delivery lead times. Consequently, many services with various prices and delivery time could be combined.
In this context, we propose a problem with a production site and several retailers facing an uncertain demand, linked by various carriers: the Stochastic One-Warehouse Multi-Retailer Problem with Lead Times (SOWMRP-LT). The objective is to minimize the average production, transportation, and storage costs as well as the unsatisfied demand.
The proposed multi-stage problem is solved approximately with a rolling horizon framework solving its two-stage version. Heuristics speed-ups were developed to cope with the long width of the window simulated by the rolling horizon, which is required by the large lead time taken into account.
10h55 - 11h20
Multi-depots concrete delivery problem with a heterogeneous fleet
We study the daily scheduling of concrete delivery. Truck drivers are dispatched based on assignment priority, and remaining working time over the horizon planning. A customer may be served from several available production plants. Due to the heterogeneity of the fleet, the number of visits at a customer location is not known ahead of time. The objective is to schedule drivers while minimizing travel times, waiting duration, idle time, and overtime. We solve this problem using a greedy randomized adaptive search procedure, and a genetic algorithm, both enhanced with a local search. Computational experiments with real data show the performance of our methods over the approach used by a local company.
11h20 - 11h45
Simultaneous Estimation of Arc Travel Times and Path Choice Models
Arc travel times and path choice models are both central to a many transport problems. The former are inputs to solve various optimization problems, and the latter are used to predict traffic flow. Both are estimated based on data, but the estimation problems are tackled separately under strong assumptions: Travel time estimation algorithms assume that vehicles use the shortest path whereas parameters of path choice models are estimated assuming arc travel times are exogenously given (e.g., some percentage of free-flow time). We propose a method to estimate arc travel times and parameters of path choice models simultaneously, hence relaxing these strong assumptions. Furthermore, our method can make use of data at different levels of granularity (aggregate observations of trip times or detailed trajectory observations). We illustrate the strong performance of our algorithm on a yellow cab dataset collected in New York city.
11h45 - 12h10
An exact solution approach for combinatorial bids generation in transportation procurement auctions under uncertainty
We address a new variant of the Bid Construction Problem (BCP) with stochastic clearing prices in combinatorial auctions for the procurement of truckload transportation service that tackles uncertainty on bids success and auctioned contracts materialization. We propose the first exact method to generate multiple non-overlapping OR bids while ensuring a non negative profit for the carrier independently of the auction outcomes and contracts materialization. Our experimental results highlight the good performance of our approach on the carrier's profit for the set of generated instances.