| Project | |
| Research Home >> |  |
Cooperative Decision Support
|
|
|
|
|
Our Past Work Cooperative Decision Support We worked on developing decision support systems and architectures, to augment the skills of the schedulers by offering an intelligent assistant capable of providing the information needed to make a good decision. Cooperative decision support involves addressing multiple objectives, and integrating multiple problem-solving approaches with the domain expertise of the decision makers. One way of enhancing cooperation between software and the decision maker is to produce a set of high quality candidate solutions, evaluated with respect to multiple business objectives. This allows decision makers to gain important insights into the tradeoffs between multiple, possibly competing objectives, without having to impose weighting factors for different objectives a priori. To implement this approach to decision support, we used an agent-based architecture called Asynchronous Team (A-Team). An A-Team is a team of software agents that cooperate to solve a problem by dynamically evolving a shared population of solutions. This architecture supports our aim of providing an intelligent assistant that can work cooperatively with schedulers to produce multiple solutions that illustrate tradeoffs. The A-Team architecture allows us to combine multiple problem-solving approaches and increases the likelihood that good solutions can be found in a reasonable amount of time. Beyond the benefits of using multiple algorithms independently, the A-Team architecture allows the agents to cooperate synergistically, and leads to higher quality and more diverse solutions. Agents may focus on optimizing different objectives and explore different parts of the search space for feasible solutions. The A-Team architecture further enhances this synergistic cooperation by enabling the scheduler to function as an agent. A successful application of this multi-criteria decision support approach is the IBM paper mill scheduling system. This system has been developed at the IBM T.J. Watson Research Center over the past couple of years and has been successfully deployed at multiple paper mills in North America. Computer World reported that within one year after starting to use our system, Madison Paper Industries saved $2 to $3 million in trim and transportation costs. Positive effects of our system on scheduling flexibility, efficiency and greater responsiveness to order changes are reported in Pulp and Paper, in addition to the improvements in cost and customer satisfaction. In 1998 the IBM paper mill scheduling suite was awarded the 1998 Daniel H. Wagner Prize for Excellence in Operations Research Practice by INFORMS, the Institute for Operations Research and the Management Sciences. Here is our paper. Please see our papers in publications section for additional details. |
||
| Privacy | Legal | Contact | IBM Home | Research Home | Project List | Research Sites | Page Contact | ||||