Design And Implementation Of The Ratbot Cyborg Intelligent Research Support System

Cyborg intelligence,which is aimed to combine the biology intelligence and the machine in-telligence,is an important research direction in the future artificial intelligence field.We propose a ratbot research support platform based on the requirement for the application of cyborg intelli-gence,which should be able to conduct a wide range of scenic experiments and make the creature interact with the machine in a real-time environment Taking a rat as the research subject,the camera array is used to capture a wide range of experimental scenes which are scalable.Then we improve the traditional image stitching pipeline and present a real-time video stitching framework using the dynamically optimal seam updating.In order to achieve the real-time interaction between the computer and the rat,we use a wireless Bluetooth stimulation backpack.Additionally,forthe application of this architecture,we use the trajectory mining technique to analyze the differences of the ratbot's performance in the maze-learning task.The results indicate that,comparing with the pure biosome,the cyborg intelligence:animal-robot has a better learning and memory ability.The detail contents are as follows:(1)Based on scalable acquisition requirements,we propose to use a camera array unit instead of a conventional single camera as a video capture device,and a formula to calculate the number of the cell in the array.In order to make this system suitable for the study of cyborg intelligent animal-robots,we also implemented a real-time video stitching algorithm based on dynamically updating the optimal seam.By using the pre-computing transformation matrix of each unstitched patch,dynamically region warping and the appropriate descriptor/blending method,we optimize the time-consuming image stitching pipeline to achieve about 23 frames per second in the video acquisition and processing steps,which satisfies the fundamental needs of the real-time closed-loop control of animal-robots.(2)As an application example of ratbot cyborg intelligent research support system proposed in this thesis,we compare the performance and the differences of the ratbot with or without MFB stimulation reward in the task of spatial learning in multi-stage T-maze.Through the traditional computer vision methods,we extract the characteristics of head,center of gravity and tail of the ratbot and construct their movement trajectories in the maze.By studying its trajectories'heat map and mining frequent patterns of its trajectory sequence,we found that the ratbot with MFB stimulation reward performs better in the maze-learning task than the ratbot without it,which also confirms that the frequent patterns we extracted can be used as indicators to measure the learning ability of ratbot in different conditions.