Didier Josselin

Vehicle routing problems with alternative paths: an application to on-demand transportation

The class of vehicle routing problems involves the optimization of freight or passenger transportation activities. These problems are generally treated via the representation of the road network as a weighted complete graph. Each arc of the graph represents the shortest route for a possible origin-destination connection. Several attributes can be defined for one arc (travel time, travel cost, etc.), but the shortest route modeled by this arc is computed according to a single criterion, generally travel time. Consequently, some alternative routes proposing a different compromise between the attributes of the arcs are discarded from the solution space. We propose to consider these alternative routes and to evaluate their impact on solution algorithms and solution values through a multigraph representation of the road network. We point out the difficulties brought by this representation for general vehicle routing problems, which drives us to introduce the so-called fixed sequence arc selection problem (FSASP). We propose a dynamic programming solution method for this problem. In the context of an on-demand transportation (ODT) problem, we then propose a simple insertion algorithm based on iterative FSASP solving and a branch-and-price exact method. Computational experiments on modified instances from the literature and on realistic data issued from an ODT system in the French Doubs Central area underline the cost savings brought by the proposed methods using the multigraph model.

Optimization of occupancy rate in dial-a-ride problems via linear fractional column generation

In this paper, we consider a dial-a-ride problem where the objective is to maximize the passenger occupancy rate. The problem arises from an on-demand transportation system developed in a rural zone in France, where the objective of encouraging people meeting is pursued. We address the solution of the problem with a column generation approach, applied to a set partitioning formulation where the objective function is fractional. Based on the literature on linear fractional programming, two methods are developed to deal with this fractional objective. Experiments permit to compare these two approaches and to evaluate the impact of the new objective compared to a standard min-cost or min-time optimization.

Optimisation using Natural Language Processing: Personalized tour recommendation for museums

This paper proposes a new method to provide personalized tour recommendation for museum visits. It combines an optimization of preference criteria of visitors with an automatic extraction of artwork importance from museum information based on Natural Language Processing using textual energy. This project includes researchers from computer and social sciences. Some results are obtained with numerical experiments. They show that our model clearly improves the satisfaction of the visitor who follows the proposed tour. This work foreshadows some interesting outcomes and applications about on-demand personalized visit of museums in a very near future.

Impact of the Norm on Optimal Locations

Our research sets in the field of k -facilities location-allocation problems. More precisely, we developed an approach, that aims to assess the sensitivity of centers to different metrics used to locate them, according to various spatial distributions of demands. We start by a concise state of the art. Then, we justify our approach with several examples on real or simulated data. Two mathematical formalisms of the 1---facility problem and the sensitivity estimation are provided. We finally propose different ways to assess the influence of points of demand on the centers, including exploratory spatial data analysis and random simulations on consequent samples.

Impact of a Change of Support on the Assessment of Biodiversity with Shannon Entropy

The research deals with the Modifiable Areal Unit Problem (MAUP). The MAUP is a common scale effect in geostatistics relating to how a studied territory is partitioned and to the ecological fallacy problem due to spatial data aggregation. We processed a biodiversity assessment using the Shannon index on a set of remote sensing data (SPOT 5) on the Ventoux Mount (Southern France). We applied the calculation on different geographical areas, with different sizes, shapes and spatial resolutions to test the effect of support change on the biodiversity measures. We proposed a method to aggregate the data at several imbricated scales so that the loss of biodiversity due to the spatial autocorrelation can be estimated separately from the MAUP. The concept of ‘pertinent’ scale is then discussed through two biodiversity criteria, a quantitative one (the Normalized Difference Vegetation Index, which evaluates the biomass quantity) and a qualitative one (a species typology, coming from a supervised classification of remote sensing data and experts maps).

An Oriented Convergent Mutation Operator for Solving a Scalable Convergent Demand Responsive Transport Problem

This paper presents a method for solving the convergence demand responsive transport problem, by using a stochastic approach based on a steady state genetic algorithm for enumerating a set of optimizing sprawling spanning trees, which constitute the best solutions to this problem. Specifically designed to speed up the convergence to optimal solutions, we introduce an oriented convergent mutation operator, allowing multi-objective considerations. So this solution lays the first stakes for considering real-time solving of such a problem. Led by computer science and geography laboratories, this study is provided with a set of experimental results evaluating the approach

Revisiting the Optimal Center Location. A Spatial Thinking Based on Robustness, Sensitivity, and Influence Analysis

In this paper we deal with different metrics using Lp norms in the 1-facility location problem and their properties. We propose to revisit the problem of optimal center location by discussing the properties of three well-known centers in 2-dimensional space: The 1-median for L1, the 1-center (Chebyshev center) for L∞ and the gravity center for L2, respectively, the median, the mean, and the center of extreme values in one dimension. The contribution of the research concerns methods to map influence and sensitivity that provide valuable and complementary information on space for decision making in territorial planning. We also discuss the center properties according to the primary objectives of equity, equality, and efficacy in the access to a facility. In a spatial-thinking approach, we present some methodological propositions to obtain robust and durable centers in geographical space, that rely on the adaptation of the general frame of the Lp norm to the planning objectives.

Quantitative analysis of pollutant emissions in the context of demand responsive transport

Nowadays, transport contributes significantly towards environmental problems (about 50% of the total CO and NOx). To date, most environmental issues due to transport have focused on general transportation methods. On the other hand, a popular approach of transport - Demand Responsive Transport (DRT) - has been studied for various aspects but rarely from the perspective of environmental issues. In this paper, we investigate the impacts of DRTs on pollutant emissions. For this purpose, we adapt a method established by European research co-operation - Methodologies for Estimation of Emissions from Transport (MEET). We create a specific model to estimate the pollution of a DRT system (GREEN-DRT) adapted from the MEET. We simulate DRT operation on three overlapping territories in France. The results show that optimising the DRT induces a significant decrease of pollutant emission due to the reduction of vehicles and travelled distances.

Relocation optimization of electric cars in one-way car-sharing systems: modeling, exact solving and heuristics algorithms

ABSTRACT Car-sharing system with electric cars is a very convenient service for urban transportation: it allows users to pick up a vehicle at a station and rent it during a short time. To manage this kind of system in the best way, it is necessary to solve the critical problem of vehicle stock imbalance across the stations. Several decision levels must be considered to balance the car distribution by taking into account the quality of service and the system operation cost. To this end, a linear programming model is proposed to formalize the problem in a mathematical framework, which allows the computation of optimal vehicle distribution strategies. To make our solution time efficient and usable for solving large problems, a greedy algorithm and a tabu search algorithm are proposed. These two algorithms are applied to the Auto Bleue network in Nice and its surrounding (France) using extensive simulations. Besides, an integrated mapping method is provided within the Geographical Information System QGIS to estimate flows and their locations. Numerical results demonstrate that the tabu search algorithm is able to find near-optimal solutions and good compromises between client satisfaction, number of staff agents and vehicles used, and computing time.

Personalized sightseeing tours: a model for visits in art museums

ABSTRACT This article describes a method to provide adapted visit tours in art museums according to the preferences expressed by the visitor and exhibits prestige. It is based on a dual approach with, on the one hand an automatic textual analysis of the official information available online (labels of exhibits) that allows to rank the exhibit attractiveness for a standard museum visitor. On the other hand, individual preferences are also taken into account to adapt the visit according to the personal cultural awareness of the visitor. We use operations research to solve a routing optimization problem, aiming at finding a visit tour with time constraints and maximization of the visitor satisfaction. Depending on the instance size and the problem scale, an integer linear programming (ILP) model and a greedy algorithm are proposed to recommend personalized visit tours and applied on two museums: ‘Musée de l’Orangerie’ in Paris and ‘National Gallery’ in London. The obtained results show that it is possible to recommend a good tour to visitors of an art museum by taking into account the common prestige of the exhibits and the individual interests, joining automatic text summarization and routing optimization in a limited geographical space.