SBIR Phase I: Evolving Object Neural Networks

这个小型企业创新研究第一阶段项目将研究在复杂环境中产生智能行为的问题。产生这种行为的需要在许多领域都得到了证明，包括游戏、商业运营和军事行动。这项研究的目的是为自适应系统确定适当的方法，以探索其他可能的资源分配方式，并学习如何在只需最少人工干预的情况下对刺激做出反应。这项研究将集中于进化计算和神经网络的结合，特别是将包括对象网络，这些对象网络检查场景的各个部分，并结合起来做出关于适当行动方案的决定。预期的结果包括确定发展这些复杂对象神经网络的适当方法，以及结果是否适合直接应用于商业市场，特别是在软件娱乐领域。可能的商业应用包括动态战斗环境的战略评估，供应链管理和物流中的商业运营的优化，以及视频游戏中智能对手的开发。自然选择公司计划将重点放在视频游戏产品上，将其作为将研究转化为商业应用的最直接方法，具有重大影响。

This Small Business Innovation Research Phase I project will investigate the problem of generating intelligent behavior in complex settings. The need for generating such behavior is evidenced in many areas, including game playing, business operations, and military engagements. The objective of the research is to identify suitable means for a self-adaptive system to explore alternative possible allocations of resources and to learn how to respond to stimuli with only minimal human intervention. The research will focus on the combination of evolutionary computation and neural networks, and, in particular, will include object networks that examine subsections of a scene and are combined to render a decision regarding the appropriate course of action. The anticipated results include the identification of appropriate methods for evolving these complex object neural networks, as well as the suitability of the results for direct application in commercial markets, particularly in the area of software entertainment The possible commercial applications include strategic assessment of dynamic combat environments, optimization of business operations in supply-chain management and logistics, and the development of intelligent opponents in video games. Natural Selection, Inc plans to focus on video game products as the most direct method for bringing the research to a commercial application with significant impact.