自主无人系统的具身认知智能框架  

Embodied cognitive intelligence framework of unmanned autonomous systems

作　　者：孙长银 穆朝絮[2,3] 柳文章 王晓 SUN Changyin;MU Chaoxu;LIU Wenzhang;WANG Xiao(Anhui University,Hefei 230039,China;Engineering Research Center of Autonomous Unmanned System Technology,Ministry of Education,Hefei 230039,China;School of Artificial Intelligence,Anhui University,Hefei 230039,China;Anhui Provincial Engineering Research Center for Unmanned System and Intelligent Technology,Hefei 230039,China)

机构地区：[1]安徽大学,合肥230039 [2]自主无人系统技术教育部工程研究中心,合肥230039 [3]安徽大学人工智能学院,合肥230039 [4]安徽省无人系统与智能技术工程研究中心,合肥230039

出　　处：《科技导报》2024年第12期157-166,共10页Science & Technology Review

基　　金：国家自然科学基金创新研究群体项目(61921004)。

摘　　要：自主无人系统是一类具有自主认知、运动规划、自主决策和推理能力的智能系统,其目标是在有限甚至没有人工参与的情况下完成复杂开放动态场景中的通用任务。针对自主无人系统在跨域协同任务上往往面临协同感知效率低、自组网通信可靠性差、资源调度流程慢、任务分配易冲突等一系列问题,探讨了融合大模型和生成式人工智能技术,构建了“大模型+自主无人系统+人工智能生成内容”为一体的自主无人系统“算-控-测”具身认知智能架构,以推动自主无人系统具身认知智能应用落地。Unmanned autonomous systems(UASs)are intelligent systems endowed with autonomous cognition,motion planning,autonomous decision making and reasoning capabilities.Their goals are designed to perform and complete common tasks in complex,open and dynamic scenarios with limited or even no human participation.In terms of the challenges UASs faced in cross-domain collaborative tasks,such as low efficiency of collaborative perception,poor reliability of Ad Hoc network communication,slow resource scheduling,and conflict-prone task allocation,this paper explored how to combine large models and generative artificial intelligence(GAI)technology to construct the“compute-control-test”embodied cognitive intelligence framework of UASs integrating“large model+autonomous unmanned systems+artificial intelligence generated content(AIGC)”.It will provide valuable reference for advancing the technological implementation and practical deployment of UASs with embodied cognitive intelligence.

关 键 词：自主无人系统 具身智能 大模型 生成式人工智能 通用智能机器 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]