Artificial neural network chip and robot with artificial neural network chip

Abstract

The invention provides an artificial neural network chip and a robot with the artificial neural network chip. The chip comprises a silicon substrate and a 3D memory array, between which a peripheral logic circuit and an artificial neural network circuit are arranged; the chip is prepared based on 3D nonvolatile memory processes; the peripheral logic circuit of a 3D nonvolatile memory and the artificial neural network circuit are realized by employing front-end processes. The robot comprises the artificial neural network chip. The artificial neural network chip employs the 3D nonvolatile memory as a mechanism of memory, possesses an ultra-large capacity data storage capability, and can still maintain data after power failure. Accordingly, compared with traditional robots, the robot with the artificial neural network chip has the characteristics of high flexibility, strong learning capability and adjustable dynamic condition.

Description

technical field [0001] The invention relates to the field of intelligent robots, in particular to an artificial neural network chip and a robot equipped with the artificial neural network chip. Background technique [0002] Since 2013, my country has become the world's largest consumer of robots. According to the statistical report of the International Federation of Robotics (IFR), the global sales volume of industrial robots was about 180,000 units that year, of which about 37,000 units were sold in China, and the market growth rate is expected to remain at more than 40% per year in the next five years. Automation and intelligence are becoming a rational choice for many enterprises' production lines, and a large number of industrial robots have begun to appear on the manufacturing line. Robots generally consist of three parts: mechanical part, sensing part, and control part. These three parts can be divided into six subsystems: drive system, mechanical system, sensory syst...

AI Technical Summary

Problems solved by technology

It can be seen that the programming languages ​​and system structures of robots from different manufacturers are different, which undoubtedly further makes the work of programmers more difficult.

Thirdly, a simple human-computer interaction system cannot precisely control the robot, such as a cleaning robot. Since the operation target often changes, it is necessary to continuously adjust the control of the robot, but the precise adjustment cannot be achieved, which makes the operation effect unsatisfactory.

Finally, the traditional robot uses CPU or MCU as the underlying control hardware. Although it has a certain storage function, it is far from realizing the self-learning function. Therefore, such a robot will not become more "smart" due to the increase in user time.

Therefore, traditional robots still need to be improved

Images