基于DBTNN算法的场景调度策略研究

构建了一种能有效描述大规模虚拟环境的场景图和基于该场景图的场景多叉树,以此为基础提出了基于改进的动态二叉树神经网络(Dynamical Binary-tree Based Neural Network,DBTNN)进行场景调度的新方法,并给出了神经网络挖掘装配过程中视点变化的规律.同时,通过该网络输出预测下一步或下几步的视点状态信息,使场景调度具有一定的容错性,提高了场景调度的实时性并使调度更加智能化.最后给出了不同场景的测试结果,表明该算法对于大规模复杂场景有非常好的优化效果.

Research on Scene Dispatch Strategy Based on DBTNN Algorithm

A new scene graph that can effectively describe larg}scale virtual environment and a scene graph based Multi Space Partition tree(MBSP) were proposed. Proposed scene dispatch strategy of Dynamical Binary-tree Based Neural Network(DBTNN) , and gave the nerve network to dig the viewpoint change rule during the assembly process. At the same time,forecasted viewpoint state of the next step or the next steps by the network output, so that the scene had a fault tolerant scheduling, improved scheduling of real-time scenes and more intelligent scheduling. Finally, the test resups of different scenes show that the algorithm for largcscale optimization of complex scenes has very good results.