基于负信息的改进多假设定位算法  

Multiple hypothesis localization using negative information

作　　者：蔡怡昕[1] 赵明国[1] 石宗英[1] 徐文立[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《清华大学学报（自然科学版）》2008年第4期495-497,501,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(60474024);清华大学教育基金资助项目(202025001)

摘　　要：为了克服多假设定位算法在特征非唯一环境中假设数量大、收敛速度慢的问题,提出了利用负信息的改进算法。改进算法首先分析"传感器未检测到环境特征"的事件为定位提供的信息,建立与传感器模型类似的负信息模型。其次,在算法流程中考虑传感器数据中未包含期望特征和未收到传感器数据的情况,在假设权重的评价中加入负信息因子,从数学形式上相当于增加一组独立传感器,为定位提供了额外的信息。仿真实验表明,该算法能够减少假设数量,加快算法收敛,并能够在特定条件下解决多假设定位算法无法唯一定位的问题。A multiple hypothesis localization （MHL） algorithm considers a number of hypotheses but converges slowly when there are non-unique features in the environment. The convergence is improved with an MHL algorithm that uses negative information. The method analyzes the information provided when sensors do not detect expected environmental features and builds a negative information model similar to the sensor model. The algorithm then adds the negative information to the evaluation of the hypotheses when there are no expected features in the sensor data or no sensor data at all, which can be viewed as adding independent sensors providing extra information for the localization. Simulations demonstrate that the improved MHL has less hypotheses while running, converges faster, and can cope with conditions for which the MHL can not obtain a unique result.

关 键 词：智能机器人 机器人定位 多假设定位 负信息 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]