1 présentation

• 
  14h00 - 15h00

  Pricing and bundling decisions considering drivers’ behaviour in crowdsourced delivery

  • Archetti Claudia, prés., University of Brescia

  Challenges in last-mile delivery such as high costs, rising customer expectations, and congested urban traffic have encouraged innovative solutions like crowdsourced delivery, where online platforms leverage the services of drivers who occasionally perform delivery tasks for compensation. A key challenge in crowdsourced delivery is that occasional drivers’ acceptance behavior towards offered tasks is uncertain and influenced by task properties and compensation amount. The current literature lacks formulations that entirely address this challenge. Hence, we formulate an optimization problem that maximizes total expected cost savings by offering bundles of tasks to occasional drivers. To this end, we simultaneously determine the optimal set of task bundles, their assignment to occasional drivers, and the corresponding compensation values for each bundle-driver pair while considering bundle and compensation dependent acceptance probabilities of occasional drivers.

  The vast number of potential task bundles, combined with incorporating occasional drivers’ acceptance probabilities via logistic functions, leads to a mixed- integer nonlinear programming (MINLP) formulation with exponentially many variables. Using mild assumptions, we address these complexities by exploiting properties of the problem, leading to a linearization of the MINLP which we solve via an exact column generation algorithm. Our algorithm considers a variant of the elementary shortest path problem with resource constraints (ESPPRC) that features a nonlinear and nonadditive objective function as its subproblem, for which we develop tailored dominance and pruning strategies.

  We introduce several heuristic and exact variants, and perform an extensive set of computational experiments evaluating the performance of the different algorithms and the analysing the structure of the solutions. The results indicate the efficiency of exact and heuristic algorithm variants for instances with up to 120 tasks and 60 drivers. The sensitivity analysis indicates that sensitivity to compensation is the most influential factor in shaping the bundle structure.