Overview of an adaptive supply chain

Three important entities are represented in the AA2 software model of an adaptive supply chain:

• Two different types of aircraft
• Field services engineers
• Depots

Each entity has different characteristics and requirements, although they must co-operate to meet the operational goals of the supply chain. The aircraft require routine servicing and occasionally breakdown - and at times there may be large-scale breakdowns. The engineers have different skills sets, locations and therefore provide different levels of service. They also have scheduled time on and off, take holidays and occasionally report in sick. The depots carry spare parts and tools, however, like any 'just in time' operation there is a fine line between over and under stocking inventory levels.

In AA2 software agents represent each of these entities. The Lost Wax Agent Framework allows the agents to maintain a constant dialogue between them. When an aircraft breakdown or requires servicing, the agent will send out a service request based on the task and service level agreement for the aircraft. All the engineer and depot agents that are able to respond do so, enabling the most appropriate resource to be allocated to the service request.

In some cases, engineer agents will collaborate because more than one skill set is required. The engineer agents invariably collaborate with depot agents, although a depot may collaborate with another depot to supply parts.

Adaptable and economic behaviour
Each response to a service request will differ depending on a number of real world variables. For example, one engineer may be physically closer to the aircraft but may need to go to a depot to get parts. Another engineer may have the right parts but needs to collaborate with another engineer because they do not have a complete skill set. Even with a simple three-entity model where each entity is constrained by relatively few operating parameters, a large number of permutations can be generated.

The agent representing the aircraft initiating the service request evaluates the responses from the engineer/depots agents, and accepts the best offer. This process continues to be dynamic. Again in the real word unexpected events occur. For example, an engineer's current job may over-run and as a result they cannot get to the next one in the agreed time frame. Depots may suddenly go offline, or there is a sudden influx of very high priority jobs. If a service providing agent realises that it unexpectedly cannot meet it's service level the agent will place the job back on the market for other agents to bid for the job again. It is this continuous process that is at heart of the 'adaptive' element of the supply chain.

Economic parameters can be used to determine the behaviour of agents. The engineer agents can be tasked with trying to achieve a specified level of utilisation. The depot agents may be operating to an economically optimal level of stock. Aircraft agents could include in their evaluation of responses a trade off between the required service level and the cost of providing the service. The key factor is that the overall framework of maintaining a level of service in the supply chain can include a number of parameters that interact with each other. Invariably there is a trade-off with one operational requirement balanced again another, which is often the role taken by expert users operating in the field. In an agent-based supply chain these interactions can be analysed better, and potentially measured in monetary terms, rather than remaining inside the heads of the individuals operating within the supply chain.

A number of large global organisations and technology providers are considering this question. A key challenge is that it is difficult to put a price tag on the benefits of adaptability, although paradoxically there is ample evidence that lack of adaptability can cost companies a great deal, both in terms of money and reputation.

The conventional approach to supply chain management, and any schedule-based activity, is centralised command and control systems based on complex scheduling algorithms. These systems assume there is perfect information and all of the variables can be factored into a calculation. Deriving the optimal schedule is simply a mathematical formula. At Lost Wax we do not dispute this is true, however, we rarely see supply chains where there is perfect information and the unexpected never occurs.

In the real word numerous events conspire to destroy the schedule. It is a problem for the largest airlines and transport companies to local restaurants running a booking system. Once a centralised schedule or a command and control system is disrupted it is usually very difficult to recover, without going back to the beginning of the planning process and starting again. Invariably it is the staff on the ground that handles the detailed operational decisions to keep the business process on track. The problem with this is the tactical decision-makers very rarely have a helicopter view of the whole supply chain and as result cannot see the overall impact of their decision - or the cost implications.

A further benefit of adaptive versus centralised systems is the opportunity to develop a system iteratively and analyse the emergent behaviour. The adaptive supply chain does not expect perfect information. The adaptive supply chain will process the available information, and within constraints, will take the best next step. This means that a simple model of supply chain can be set up with a limited number of operational parameters and goals and the results analysed. Incrementally adding more parameters and goals can increase the complexity and sophistication of the supply chain. The resulting model can yield surprising insights in to the way the supply a chain operates and adapts as various disruptive events occur. Analysing every possible situation in order to calculate an optimal schedule, with every possible outcome catered for, is usually beyond the capabilities of most organisations and software developers.