基于场景-拓扑的多机器人协作环境搜索研究

Cooperative explorations in unknown environment by multi-robot have such advantages as parallelism, flexibility and redundancy. It also bring about the novel challenges including limit communication, local map merging as well as the multi-targets allocation strategy, those affect the exploration efficiency greatly. In order to handle the bottleneck problems of data registration in local map merging, arround a novel proposed scene-topological map, the temporal-spatial invariant features based on the fused multi-source information was extracted, the method of environment modeling for unstructured environment with hierarchical architecture and feature reduction, the method of cooperative map merging based on scene-topology was reserached to handle the scientific problems such as the environmental recognition based on the multi-source,temporal-spatial features,the feature matching and data reduction for huge data,to explore the driving mechanism for information interaction under the limit communication as well as the local map merging based on the hidden markov model.This project will supply the important theory and technology support to the application of disaster rescue, military reconnaissance, enemy searching and exploration in unknown environments. Therefore, it will naturally bring about huge social and economic benefits.

利用多个机器人协作探索未知环境具有并行、柔韧和信息冗余的优点，同时也带来了新的挑战：有限的通讯能力，局部地图拼接和多任务分配策略是影响多机器人协作环境搜索效率的主要因素。本项目针对多机器人协作环境搜索中局部地图拼接的数据配准瓶颈问题，围绕一种新型的场景-拓扑地图开展基于多源信息融合的时空不变特征抽取，面向非结构化复杂环境的的层次化环境建模与特征约简以及基于场景-拓扑的多机器人协作地图拼接方法研究。解决移动机器人基于多源时空不变特征的环境认知、面向海量特征集的数据约简与特征匹配科学难题，探索有限通讯能力下多机器人信息交互的驱动机制和基于隐马尔可夫模型的局部地图拼接方法。项目研究将为实际多机器人在民用灾难环境救援、军用的侦察、巡逻与搜捕以及未知环境探索等领域中的应用提供重要的理论和技术支撑，带来巨大的社会效益与经济效益。

多机器人协作未知环境探索作为多机器人协同任务执行的基本问题，与单机器人相比具有并行、柔韧和信息冗余的优点，同时也面临有限通讯，局部地图融合和多任务分配策略的挑战。本项目重点研究了地面移动多机器人系统在未知、复杂环境下的时空不变特征抽取与匹配方法;基于时空不变特征的层次化环境建模方法；基于时空不变特征的局部地图拼接方法，以及协同探索中的多任务分配与行为规划方法。取得的主要创新点包括：（1）二进制时空不变特征提取与特征匹配一致性评估方法；（2）基于时空不变特征的层次化场景-拓扑环境建模方法；（3）基于视觉词汇树与协作隐马尔可夫模型的协作地图拼接方法；（4）基于信息排斥素蚁群算法的多任务分配方法。本项目成果解决了卫星定位失效以及有限通讯带宽情况下的数据配准难题，将为实际多机器人在民用灾难环境救援、军用侦察、巡逻与搜捕以及未知环境探索等领域中的应用提供重要的理论和技术支撑，带来巨大的社会效益与经济效益。