by Tamás Máhr and Alfons Salden
How can agent-based technology reduce transport costs, traffic jams and carbon dioxide emissions? DEAL, a Dutch research project on multi-agent-based logistic monitoring and planning, is studying these questions.
The research and industry partners in the Distributed Engine for Advanced Logistics (DEAL) project are Almende BV, RSM Erasmus University, Vrije Universiteit Amsterdam, CWI, CarrierWeb, Post Kogeko and Vos Logistics. The goal of this consortium is to apply new knowledge to a broad range of products for national and international logistic services. In particular, it aims to speed up negotiations amongst software agents representing different stakeholders in dynamic logistic problems. This would enable an increasing number of logistic conditions and alternative logistic plans to be examined, and therewith a higher overall optimal performance of the logistic system to be achieved.
The current best practice for planning in a logistics company consists of human planners interacting with an order-assignment database and a tool for calculating optimal routes. The most advanced logistics tools focus on balancing well-defined goals such as minimizing the number of vehicles, the distance covered, and the number of empty kilometres driven. The task of the human planners is to make choices, often by imposing additional constraints on the optimisation tool, such that all involved parties are more or less satisfied with the schedule. This process is time-consuming, and a schedule may still be sub-optimal when the time comes for it to be executed.
Another issue is that the extra constraints imposed by human planners on the optimisation tool consist of a large number of rules that express the preferences of all the involved parties. To learn all the rules is a lengthy process and makes the logistics companies vulnerable to changes in personnel.
Outperforming the Optimum
The DEAL project develops a multi-agent based logistic planner and monitoring tool that allows planning and execution to overlap. Truck and container agents dynamically change plans to react to unexpected events by using different market mechanisms such as auctioning and de-commitment. The agents continuously search for better solutions while always maintaining an executable plan. Even while some parts of a plan are being executed, the agents keep searching for better solutions for the rest of the work, and if any incidents occur, they adjust their plans immediately.
Another issue studied in the DEAL project is the evaluation of logistics performance. It is argued that simply evaluating one of the traditional performance indicators is insufficient for a modern logistics provider. Rather, monitoring of key performance indicators of all the stakeholders is essential for successful business relations. This means that the optimisation goals of the agents are manifold. To decrease the number of optimisation goals, we introduced an evaluation framework based on fuzzy logic. This translates the values of several key performance indicators into satisfaction measures of the main stakeholder groups: management, customers, employees and society. Given these measures, agents can concentrate on generating a solution that balances the satisfaction of these four groups.
In conclusion, we see the problem faced by logistics companies to be not merely an optimisation for cost problem, but a balancing problem within an uncertain and unstable environment. Companies must provide plans for transportation vehicles such that all parties are satisfied (including management investments), while constantly facing incidents that force changes of plan.
Tamás Máhr and Alfons Salden
Almende, Rotterdam, The Netherlands
Tel: + 31 10 4049444
E-mail: tamasalmende.com, alfonsalmende.com