Research And Design On Application-Specific Processor Chip For Agricultural Internet Of Things

As the foundation of information industry,integrated circuits bear on the development of the whole economy and society system.With the further development of microcontroller technology,its application field has now been expanded to the agricultural field.However,the current microcontrollers are designed according to the standard of industry.It is urgent to study and design processor chips for agricultural applications because of the more complicated application environment of agriculture.While meeting the needs of agricultural applications,the microcontroller technology in the agricultural field needs be controlled independently to ensure agricultural information security.In view of this,in this paper an agricultural application-specific processor chip is designed and implemented.The main research results are as follows:(1)The application mode and overall design of the agricultural application-specific processor are proposed.In terms of five application scenarios of agricultural internet of things,such as field crops,facility horticulture,aquaculture,livestock husbandry,and comprehensive agricultural types,as well as agricultural planting,breeding,cultivation of improved varieties,storage of agricultural products,agricultural mechanization,classification and processing of agricultural products,and agricultural meteorology and environment,these typical application patterns of the agricultural application-specific processor chip for smart agriculture have been analyzed,and the parameter specifications of the chip and the requirements of the acceleration module for image processing are obtained.(2)An architecture of Agricultural Application-Specific Processor(AASP 1.0)is proposed based on the CISC instruction set and Harvard architecture.Using the technology of IP reuse,the AASP SoC system is implemented,mainly including a CPU and 5 peripheral IPs(ring oscillator,PLL,crystal oscillator,8-channel ADC and LDO).(3)A SoC chip of agricultural application-specific processor is implemented,and the RTL design,simulation verification,layout design and tapeout verification are carried out.The application-specific processor chip for agricultural Internet of Things mainly includes 32 GPIO ports,8-channel AD,DAC through PWM,firmware storage using built-in ROM and external FLASH.Mixed Signal technique based on SMIC 0.18?m for tapeout is used.(4)The test of application-specific processor chip for agricultural Internet of Things AASP 1.0 is completed and its test circuit and application circuit are designed and implemented.The power consumption,serial communication,application programming,general-purpose I/O ports and other functions are tested,and the application circuit system is implemented to verify these functions and performance of the chip.The test results show that the chip meets the requirements of the design.(5)The application of lightweight deep neural network for image processing under the application mode of the application-specific processor chip for agricultural Internet of Things is presented.In order to realize the functions of image processing for information acquisition,storage,transmission,processing and analysis and information security,the computational volume and storage space of a lightweight deep neural network model named SqueezeNet are analyzed.During training,the model was fine-tuned through small dataset and was successfully transferred to agricultural image classification.In conclusion,in the process of AASP1.0 hardware implementation,in terms of RTL source code,it can add different modules,such as security,image processing and other expansion modules,and can be flexibly configured storage capacity.Because of scalability and flexibility,the methods referred to may ultimately help to achieve an independent and controllable technology,and ensure the realization of information security.