Collaborative framework of an intelligent agent system for efficient logistics transport planning

- We address the barge planning issue in hinterland transport domain.
- We define and formulate the planning model for each participant.
- We use hybrid algorithm to assist planning process.
- We use agent system to support information exchange and decision making.
- The result is validated by means of case and comparative study.

In modern logistics chain, planning system which been applied for supporting daily operations, is becoming more information-intensive and technologically-dependent. Due to the growing operational complexities, the planning system is not only required to focus on ensuring the feasibility of daily plan, but also be capable of manipulating conflict of interest (COI) by exploiting efficient negotiation. This paper addresses a challenge of hinterland transport planning caused by limited information sharing, lack of collaboration and COI. We design collaboration and decision-making mechanisms for the implementation of autonomous control by means of multi agent technology and hybrid heuristics. It aims at providing quality plan to achieve high level of performance and robustness in hinterland logistics. From a system point of view, a service oriented architecture is proposed to integrate agent paradigm with a web-based planning system. From an operational point of view, special attentions are paid on establishing multi-dimensional collaborations between different stakeholders under the support of game theoretic approach. Specifically, we define how planning activities are executed, which level of collaborations are incorporated and how benefits are achieved with a global point of view. Due to the NP-hard nature of the addressed problem, an integrated NGSA-II and simulated annealing algorithm is implemented for assisting decision making. The parameters associated with the algorithm are tuned based on the Taguchi method. Case and comparative study will be presented to validate the appropriateness and performance of the approach.