Operational impact of in-advance travel requests dispatch optimization in Dynamic Demand Responsive Transportation
Louis Zigrand  1@  , Roberto Wolfler Calvo  2@  , Emiliano Traversi  3@  , Pegah Alizadeh  4@  
1 : Laboratoire d'Informatique de Paris-Nord
LIPN, UMR CNRS 7030 - Institut Galilée – Université Paris 13
2 : LIPN - Université Paris Nord
université Paris 13
3 : Laboratoire d'Informatique de Paris-Nord  (LIPN)
Université Paris XIII - Paris Nord, CNRS : UMR7030, Institut Galilée
Institut Galilée 99, avenue J.B Clément 93430 VILLETANEUSE -  France
4 : Laboratoire d'Informatique de Paris-Nord
LIPN, UMR CNRS 7030 - Institut Galilée – Université Paris 13

This work shows the potential of a re-optimization of the planning of the vehicles in a Dynamic Demand Responsive Transport service before they leave their depots under the hypothesis that the operator is able to forecast the requests that will arrive during the day. The type of problem we are dealing with imposes to use a new objective that maximizes the expected acceptance rate of future requests, since the requests scheduled in-advance are mandatory. This objective function is incorporated in a Combinatorial Benders Decomposition to solve the Dynamic Dial-a-Ride Problem. A key ingredient for the success of this solver is the use of a new clustering-based initialization of the Master Problem. The computational results show the effectiveness of this approach when applied to instances extracted from actual services provided by Padam Mobility, an international company working in Shared Mobility Systems. The proposed method provides a substantial performance gain though traditional heuristics are kept to manage online insertions afterwards.


Personnes connectées : 150 Vie privée
Chargement...