基于ACT-R的认知建模方法在人-机器人空间认知交互中的应用  

Application of Cognitive Modeling Method Based on ACT-R in Human-Robot Spatial Cognitive Interaction

作　　者：牟献梁 王春慧[1] 田雨[1] 谭丽芬[1] 张绍尧 

机构地区：[1]中国航天员科研训练中心人因工程重点实验室,北京100094

出　　处：《航天医学与医学工程》2017年第2期135-142,共8页Space Medicine & Medical Engineering

基　　金：试验技术青年基金(2014SY54C0702);人因工程重点实验室基金(SYFD140051802;SYFD1400618);装备预研基金(9140A26070214KG57431)

摘　　要：目的研究空间认知任务中认知建模方法在提高人-机器人交互自然度和效率中的作用。方法分析可计算认知地图的表征方法,通过人-人交互实验研究了典型空间认知任务中人类空间语言的表达规则、空间参考框架选用规律以及人类视角变换能力对任务绩效的影响;采用ACT-R认知结构表征人类可计算的空间认知地图,为机器人建立拟人的认知结构体系,使机器人能够以人的视角和语言特征与人交互。结果在测试设定的多种工况下认知模型能够准确判定人类想要获取的目标,人-机器人交互的自然度提高,交互效率提高。结论空间认知任务中,认知建模方法有助于提高人-机器人交互自然度和效率。Objective To study the effects of cognitive modeling method on improving the naturalness and effi- ciency of human-robot interaction in the spatial cognition task. Methods How to represent a computable cogni- tion map was analyzed firstly in this paper. Then a human-human interaction experiment was conducted to get the expression rules of human＇ s spatial language, the law of choosing spatial reference frames and the influ- ence of human＇ s perspective-taking ability on spatial cognition task＇ s performance. Based on ACT-R cogni- tive architecture, a computable cognition map was represented and an anthropopathic cognitive architecture was built, which could enable the robot to interact with people on the perspective of people using natural lan- guage. Results The model could identify the target that people want to get. The naturalness and efficiency of human-robot interaction were improved. Conclusion In the spatial cognition task, cognitive modeling method can help to improve the naturalness and efficiency of human-robot interaction.

关 键 词：空间认知 自然交互 认知建模 ACT-R 视角变换 

分 类 号：R857.1[医药卫生—航空、航天与航海医学] TP39[医药卫生—临床医学]