SIEMENS

At present, customers have to wait weeks, if not months, before they can receive a new car. One reason for this delay is the large range of options. In the case of some vehicle models there are millions of possible combinations of optional equipment. The individual parts are ordered through logistics networks that have to be seamlessly integrated so that everything is available at the assembly line when a car is put together.

The people who are responsible for these networks are logistics planners. There are about 100 of them at each automaker, and each one coordinates up to 100 more planners working at supplier locations. Communicating by phone and e-mail often takes a long time, which contributes to the industry’s inability to deliver cars within five days after receiving an order, an ideal promoted by the European Union in its “Intelligent Logistics for Innovative Product Technologies” (ILIPT) research project.

The project brings together 30 organizations from the research community and the auto industry, including Daimler, BMW, Continental, ThyssenKrupp, MAN, and Hella, which have developed methods to accelerate the planning of supply chains. The centerpiece of the project is a set of market-based processes and technologies for software agents that are being developed by Siemens Corporate Technology (CT).

Software agents are autonomous programs that operate independently and respond to altered conditions — ideally almost the way a person does, but much more quickly. They are already being used in robots, and in the future they will also take on the task of performing negotiations in virtual marketplaces. The idea behind ILIPT is that manufacturers and suppliers in their branching supply networks will automatically coordinate their output quantities and delivery dates through the Internet using software agents, without violating the terms and conditions they have agreed on in their contracts. This doesn’t put logistics planners out of work; they still have to negotiate the basic contracts and configure the negotiating strategies of the agents.

Lightning Logistics. These intelligent computer programs can even take over price negotiations and perform them recursively in a cascading fashion, in a manner analogous to the structure of the supply chain. This means that the demand for parts that is predicted by an automaker is negotiated with the supplier, that supplier negotiates with its own suppliers, and so forth. If two agents in this chain cannot reach an agreement, they negotiate backwards again in order to meet all of the specified parameters. The agents also take part in auctions if multiple suppliers are competing for the same contract.

To ensure that agents know how much elbow room they have when it comes to negotiations, they are provided with capacity limits as input parameters. The supplier, for example, states its production capacities for certain modules. It takes into account such constraints as plant vacations, retooling times, and even strikes. If a supplier’s warehouse catches fire, the agents can re-plan the supply chain — “in a matter of seconds,” says Jan-Gregor Fischer, a CT expert for agent technologies in Siemens’ Intelligent Systems and Control Global Technology Field.

The availability of parts determines their price. For example, prices rise if parts become scarce. To make bids in the virtual marketplace, the supplier establishes minimum and maximum values that its agent must abide by in order for the transaction to be profitable. The same is true for the purchaser. These input parameters are secret, of course — as they are today when a vendor and a purchaser sit across from one another at the negotiating table.

But there are cases when demand forecasts cannot be brought into agreement with delivery capacities or the prices expected by participants. In this case, agents initiate a second type of negotiation in which they renegotiate minimum and maximum limits. Here as well, there are quantity and price constraints that the agents are required to heed.

High Gear Supply Chain. Agent technologies open up completely new possibilities. For example, customers will have many more options when choosing equipment for a dream car, but they will be able to take delivery in just a few days. If a customer orders his vehicle through the Internet, the required parts can be reserved immediately according to the “build to order” principle. They are then delivered to the production line without delay and assembled. This is not possible with the forecasting and long-term supply contracts commonly used today, which involve high warehousing expenses and require that manufacturers anticipate demand for parts months in advance.

The supply chain of the future will have to adapt dynamically to demand. This is becoming all the more important because the special planning for individual model series no longer reflects the state of the art. The future belongs to module strategy, which puts similar components into different models — but these additional dependencies can make logistics more complex. “Software agents can help to make these complex scenarios manageable,” says Thomas Sommer-Dittrich, who is responsible for production management at Daimler Research in Ulm.

Agents Hit the Road. The ILIPT project launched in 2004 has now been brought to a close. The concepts that were developed for collaborative capacity planning were integrated into a software demonstrator designed by Daimler and Siemens. Siemens CT provided D’ACCORD, a market-based negotiation component, and Daimler implemented the software agents. The partners took the demonstrator on a road show during which they showed off the benefits of these new technologies to the European auto industry.

Daimler is already using knowledge gained from the project — such as software for network analysis — but is not yet employing agents as standard tools for logistics. In addition to unresolved technical questions, such as how to create a secure infrastructure for the agent-based systems among multiple manufacturers, an important prerequisite for this approach will be an increase in the modularization and standardization of components, particularly across multiple brands and manufacturers, so that a market for standard bids can be formed.

“Another aspect that still has to be resolved is how the sometimes very extensive certifications of products and processes will occur in the new framework,” says Sommer- Dittrich. “These aren’t things that can be done ahead of time.”

Only when these issues have been resolved will agents be able to negotiate in marketplaces that include multiple companies. “That’s when agents will bring their full value to bear and help cut costs for everyone involved,” says Sommer-Dittrich.