Development of a Cognitive Robot Powered by the Autonomous Learning Engine

认知机器人（CR）是一种能够模仿人脑自主学习和推理的类人机器人技术。传统命令式机器人的智能能力是基于预加载的存储程序控制技术。该研究计划将命令式机器人扩展到由认知计算和自主学习技术驱动的CR。短期目标是建立一个新的CR和自主机器人学习技术的发展。我们的长期目标是使技术成熟到一种新型的认知计算机，并通过我们实验室的合作伙伴，如IBM，Oracle，AAI，iCore和TRLabs，将先进的CR技术转移到行业中。 该研究计划分为三个连贯的阶段。第一阶段设计了认知反应器的认知结构和自主功能，并通过概念代数对其进行了严格的建模。第二阶段开发CR的自主学习引擎，使用一种新颖的概念，句子和文章自下而上的学习递归过程。第三阶段在工业平台上建造CR，如本田的Asimo和AAI的e-Puck机器人。CR将通过系统测试和模拟进行确认。试点研究表明，在实施公约与建议方面取得了令人鼓舞的成果。 这项研究的新奇在于开发一种新型的CR，将当前的命令式机器人扩展到具有认知学习能力的自主机器人。这项研究不仅导致CR在工业中的应用成熟，而且还导致认知计算机，机器人和指称数学的理论进步。这项研究的预期影响是CRs可以通过模仿大脑来学习。他们还可以将学到的知识传递给具有兼容知识模型的对等CR。CRs将增强广泛的工业应用，如制造自动化系统、无人空间系统、软件代理、语义搜索引擎、先进的医疗保健系统和智能家居设施。

A cognitive robot (CR) is an emerging technology of humanoid robots that is capable of autonomous learning and reasoning mimicking the human brain. The intelligent abilities of traditional imperative robots are based on preloaded stored-program-control technologies. This research program extends imperative robots to a CR powered by cognitive computing and autonomous learning technologies. The short-term objectives are to build a novel CR and to develop the technologies for autonomous robot learning. The long-term objectives are to mature the technology towards a novel cognitive computer and to transfer advanced CR technologies to the industries via partners of my laboratory such as IBM, Oracle, AAI, iCore, and TRLabs. This research program is organized in three coherent phases. Phase 1 designs the cognitive architecture and autonomous functions of CR, which are rigorously modeled by concept algebra. Phase 2 develops the autonomous learning engine of CR using a novel recursive process of concept, sentence, and article learning from the bottom up. Phase 3 builds the CR on an industrial platform such as Honda's Asimo and AAI's e-Puck robots. The CR will be validated by system testing and simulations. Pilot studies have demonstrated encouraging results towards the implementation of the CR. The novelty of this research is to develop a new type of CR that extends the current imperative robots to autonomous ones with cognitive learning ability. This research leads not only to the maturity of CR applications in the industry, but also to theoretical advances in cognitive computers, robotics, and denotational mathematics. The anticipated impact of this research is that CRs can learn by mimicking the brain. They can also transfer learnt knowledge to peer CRs with compatible knowledge models. CRs will enhance a wide range of industrial applications such as manufacturing automation systems, unmanned space systems, software agents, semantic search engines, advanced health-care systems, and smart home facilities.