MEMS Bionic Individual Based On Prior Knowledge, Real-time Situation Awareness And Interactive Learning

Intelligence technology characterized by sensor applications has been the center ofattention in the information age. With rapid progress of micro-nano technology, sensortechnology and information science, Micro-Electric-Mechanic Systems (MEMS) havebecome more sophisticated and can be widely applied. MEMS have the characteristicsof computers, as well as those of measurement and control instruments. However, atpresent, MEMS appear very clumsy in many respects; they fail to perform tasks whichare very simple for animals, from insects to humans. There is an urgent need tointroduce bionic intelligence to the given MEMS hardware so as to enhance their value.Living creatures’ mechanisms are extremely precise and complete, and thesemechanisms have been providing important inspiration for the development of humanscience and technology. Intelligence is the unique attribute of life, and toproduce artificial systems which can respond in a biological manner or behave likenatural life is the common goal of AI research and Artificial Life (AL) research.MEMS share similarities with biological individuals in senses, computing power, smallsize and possessing independent bodies. MEMS are the most suitable hardwareplatforms for carrying out AI research and AL research. Carrying out AL researchon MEMS can help people build a deep understanding of life phenomena, while AIresearch on MEMS can solve some of the difficulties and crises encountered by AI.Taking the bionic situational awareness and prior knowledge learning as entrypoints, this research explores the "Intelligence-as-it-might-be" and the"Life-as-it-might-be" based on MEMS hardware. Two basic dynamic principles impliedin life phenomena, information driving and time driving, are proposed. Aninformation flow model and a timeline model of biological individuals are establishedto build a bionic individual, which is of external-presented life features andinternal-running life mechanisms. The mechanisms and characteristics of animals usingtheir sensors are explored and then tested and verified by constructing bionic sensoryorgans using micro-nano sensors. With a finite state machine and a soft controller asits core elements, a MEMS Bionic Individual (MEMS BI) is constructed using acomputer program. In the MEMS BI, prior knowledge is integrated to achieve bionicsituational awareness and Bionic Learning. The Biological Life Cycle is imitated to implement and test the heredity and evolution of MEMS BI.Tests show that the bionic usage of sensors can effectively mine the informationcontained in measurements of sensors; multi-sensor information fusion can remove theuncertainty of observation and, therefore, improve the accuracy and reliability ofobservation. Tests reveal that the individual bionic intelligent system has the ability tosense its environment and to solve problems; it has the abilities of perception, cognition,learning, genetics and evolution, i.e. it is characterized by intelligent features andexhibits the behaviors of natural life. This research extracts the basic dynamics underbiological phenomena to understand life, and reproduces life phenomena on the MEMShardware.This research explores the theoretical models, generating methods andimplementation technologies of artificial life especially bionic individuals. Ultimatelyan artificial life with a variety of life features and behaving similarly to biologicalbehavior is constructed. This research tries to understand life and explore the law of life,and to explain and simulate biological individuals’ intelligence and its law in an attemptto "inject" intelligence into MEMS so as to enhance MEMS’ utility value.