Reporter ｜ Wu Xingling
Exhibition ｜ CSDN (ID: CSDNnews)
According to IDC predictions, there will be 5.5 trillion US dollars of investment in digital transformation in the world in 2021. Enterprise digital transformation is becoming the core strategy of many Chinese companies. According to the "Data and Storage Development Research Report 2019" jointly released by IDC and Inspur, China's digital transformation IT spending in 2019 exceeded non-digital transformation IT spending for the first time, accounting for 51%. %.
However, in the process of practice, some traditional industries have brought certain challenges to enterprises and developers due to the low digital information and little accumulation in related industries. How to use digital methods to help improve enterprises to increase labor productivity, find new ways of collaboration, and expand business scope?
In this regard, CSDN (ID: CSDNnews) interviewed eight industry leaders and technical experts in manufacturing, media, agriculture, logistics, and discussed the practical cases and thinking of digital transformation:
AI + IoT, creating a smart new agriculture
Now that we attach great importance to the construction of digital agriculture and rural areas, how can we empower traditional agriculture? Vigorously promote "Internet +" modern agriculture and accelerate the development of digital agriculture? Guangxi Huiyun Information Technology Co., Ltd. (hereinafter referred to as "Huiyun Information"), which has been focusing on agricultural informatization for 8 years, gave the following answer sheet-
At the beginning of the exploration, Huiyun Information encountered many difficulties and challenges:
First, agriculture is a traditional industry. The digital transformation of agriculture has not accumulated much in the past 5 to 10 years. In particular, the domestic economy is dominated by small farmers, which is different from large farms abroad. Therefore, whether it is the user's needs, or the product model, business model, Huiyun Information is exploring and stepping on the pit step by step.
The second is that agricultural informatization is relatively weak, and ordinary farmers have insufficient experience in using smartphones and information systems. If you adopt the traditional ideas of industrial informatization and service industry informatization to make software and systems, what you may make may not be suitable for farmers to use. How to make something that farmers are willing to use and convenient to use? Huiyun Information needs to think deeply.
The third is the complexity of agricultural production. Crop growth is good or bad, and there are many influencing factors. In the past, a lot of agriculture was based on human experience and judgment to make decisions and implementation, not to mention data and structure. How does technology simulate the way people think? Handling variables of various complex conditions? We allow machines to think like people, and then we can make digital products that really help people, or replace people.
Fourth, the agricultural production environment is relatively remote. When deploying hardware and other infrastructure, it is necessary to comprehensively consider the simplicity of equipment deployment, the convenience of maintenance, the stability of operation, and the ability to fight extreme weather outdoors 24 hours a day.
In response to the above challenges, Huiyun Information started from the following aspects:
First, since smart agriculture is part of the industrial Internet, to do a good job of the industrial Internet requires a certain degree of understanding of the industry. In the past ten years of intensive cultivation in the field of smart agriculture, Huiyun Information has adjusted its product strategy as it understands the needs of users: from the earliest pure agricultural information management software, to the use of Internet of Things technology to automate execution, and then To the use of artificial intelligence to realize data-based decision-making.
Second, the decision-making model used to be based on people. Now, artificial intelligence technology is used to make some products that use machines to replace people to make decisions. These products can simulate people's thinking and replace traditional experts to help users make decisions. system.
Third, in terms of hardware and infrastructure, with the development of new technologies in the Internet of Things, whether it is network technology or energy supply technology, low-power, completely wirelessly connected hardware devices can be made. During the deployment process, no network cables and wires are needed, which is convenient for installation and maintenance and suitable for agricultural production.
With the solution, how to choose the technology and platform to be implemented?
First, in the Internet of Things technology, Huiyun Information completely wrote all the code on its own at the beginning. It may not have particularly detailed considerations on stability and security, and it also has certain restrictions on the compatibility of the device.
After Huiyun Information adopted the Microsoft-based IoT Hub, it greatly improved the compatibility of the device and the stability of the device's operation: before it may only support 100,000-level sensors, and through the IoT Hub, it can support millions of sensors At the same time, the online stable operation greatly speeds up the expansion of new third-party sensor types.
In addition, because the agricultural production site often encounters network instability, how can the device control model operate normally in a networkless environment? This is a major technical difficulty. Huiyun Information uses Azure IoT Edge. As a result, gateway devices also have edge computing capabilities to solve the problem of AI models running offline.
Secondly, in terms of artificial intelligence, the efficiency of Huiyun Information in hyperparameter tuning, algorithm selection, and model judgment is improved through Microsoft Azure Machine Learning service. Only super-parameter tuning improves the efficiency by nearly 50%.
In addition, the use of AutoML automatic optimization algorithm to select the most suitable algorithm to build the model can shorten the work that Huiyun Information previously required to complete in a week to two or three days.
In addition, Huiyun Information makes it easier to unify model standards under different computing frameworks through ONNX (Open Neural Network Exchange), thereby better supporting cross-platform application of models.
With the acceleration of digital transformation, Huiyun Information has a deeper experience about smart agriculture:
In the past, smart agriculture focused more on the implementation level, using machines to replace people to work in the ground, such as automated irrigation, drone spraying, etc., mainly based on IoT technology.
But from the perspective of the complete smart agriculture chain, in addition to the implementation level, there is also a very important decision level:
Although people don't need to irrigate or spray medicine in the field now, when and how much water will they irrigate? What kind of diseases and insect pests should be controlled, what kind of medicine need to be administered, and when to apply medicine ... At present, those who need to make decisions are made by experienced people.
Although a certain degree of automation has been implemented at the executive level, if the decision-making level is always people-centered, it will cause the entire agricultural production to vary from person to person-each person ’s decision is different, making it difficult to achieve true standardized agricultural production. It is difficult to expand agricultural technical services.
The next step in the development of smart agriculture must incorporate data science, turning the traditional human-centric decision-making model based on human experience into data-centric. Through scientific data analysis, data processing and data decision-making, make more objective and easier to replicate data-based decisions.
In this way, the integration of data science must introduce AI technology. Therefore, next, the development of smart agriculture is to advance decision-making and execution. Not only the Internet of Things technology, but also must be combined with AI technology to become AIoT . This is an important direction for the future development of smart agriculture. Focused- not only to solve the problem of automation at the executive level, but also to solve the problem of data at the decision level.
AI testing, quality traceability, and reform of manufacturing quality control!
Shanghai Hongpu Information Technology Co., Ltd. (hereinafter referred to as "Hongpu Information") is a company that provides artificial intelligence + manufacturing solutions. From entrepreneurship to the present, witness the process of digital transformation of manufacturing:
At the beginning of entrepreneurship, Hongpu Information hoped to use data analysis and artificial intelligence technology to help enterprises, but because digital transformation has not become a trend, Hongpu Information encountered many challenges in its exploration : First, the degree of digitalization of enterprises is generally limited. Although most of them claim to have "mass data", most of them cannot be adopted; secondly, some companies have insufficient knowledge of AI or data analysis itself, and the communication cost is relatively high; thirdly, the data analysis solution itself has a large Uncertainty, different customer conditions, and different customer scenarios may cause the solution to be unsatisfactory , so it was difficult for data products to be implemented at the time.
" This industry has not grown to a state where solutions can be easily completed, and we have also suffered a lot. " Hong Pu Information Liu Longze recalled.
Nowadays, as the industry pays more and more attention to data analysis and digital transformation, more participants join the various links of the industrial chain to complete the entire industrial chain, and Hong Pu Information can simply complete the best work, thus Solve problems for users:
In manufacturing, quality control is an extremely important part. In the past, manual quality inspections by quality control personnel resulted in high labor costs.
To this end, Hong Pu Information launched a photovoltaic EL testing application for quality testing, saving millions of labor costs per workshop per year.It also uses technologies such as zero defect management and quality traceability to significantly improve efficiency . With the help of AI technology, Hongpu Information recreated the quality management system in production and changed the process of production management. Not only that, Hongpu Information effectively uses algorithm-assisted recommendation technology , which extends from quality management to intelligent recommendation of device parameters, saving companies about 15 million yuan in costs every year.
The most important thing is that this is not only used in individual cases. Hongpu 's solution can be copied to any enterprise in the manufacturing industry , so that every user can benefit from this solution system. In this way, enterprises can take advantage of the digital transformation to truly realize the management model innovation and produce revolutionary changes in competitiveness .
Now Hongpu Information has become a partner of Microsoft IP Co-sell, and the photovoltaic EL detection application has also entered the Azure application market, building and serving customers on the Azure cloud. Not only that, Hongpu Information also uses Azure cloud hosting, Traffic Manager, GPU training services, AutoML, Blob storage services and other technical capabilities to empower its own products and optimize the product experience. The flood of information Park all the training images are stored in memory based on Blob storage and transport, in order to achieve agile cloud development, additionally, data tagging, data transfer heavy work is also greatly simplified. This can improve the efficiency of the user's iterative algorithm model to a certain extent, which can be directly related to production efficiency!
Finally, on the hot topic of 5G in time, Guan Zhen, chief technology consultant of Microsoft (China), believes that 5G's high bandwidth, low latency, ultra-reliability, and density can greatly benefit industrial interconnection. Nowadays, in 4G scenarios, we are accustomed to data collection, uploading, computing, and reasoning are all fixed serial routes and architectures, which can be compared to a two-dimensional space; in the future we will be in a high-density, point-to-point chain Under the 5G network with very reliable roads and strong mobility, the conventional design can no longer follow the original serial route, but can be replaced by a three-dimensional three-dimensional space connecting neurons and neurons. The process of fission, re-connection, and the creation of new dimensions of data sets, the imaginable space becomes huge.
In addition to improving their ability to produce actual products through advanced industrial Internet solutions, industrial enterprises have also produced a new industrial product such as "data" in the process, which is the driving force and future of industrial enterprises' digitization.
Say goodbye to repeated high-intensity labor? Logistics robot is here!
Talking about the original intention of Beijing Jizhijia Technology Co., Ltd. (hereinafter referred to as "Jizhijia") , Shen Mu mentioned three points: First, logistics is an industry with repeatability, high labor intensity, and low added value. It is also important Infrastructure is in urgent need of intelligent transformation.
Second, the labor shortage will bring huge challenges to the logistics industry. According to the Deloitte Talent Report, the cost of labor in China has increased fivefold in the past decade. With the aging of the population and the disappearance of the demographic dividend, labor costs will continue to rise, and rising labor costs will bring tremendous pressure to labor-intensive supply chains. At the same time, the younger generation is increasingly reluctant to engage in repetitive and heavy work in the warehouse. The problems of an aging population, high labor costs, and difficulty in recruiting workers will make the logistics industry face a severe labor shortage challenge. The logistics industry urgently needs intelligent and automated transformation.
Third, unpredictable changes in market demand and business fluctuations will force enterprises to become more flexible in their supply chains. With more and more personalized demands and shorter and shorter product life cycles, companies need to respond to the needs of end consumers faster and more agilely.
Therefore, Jizhijia takes intelligent logistics as the starting point, uses big data, cloud computing and artificial intelligence technologies, and uses "Robot Internet +" to improve the efficiency of logistics operations.
In the process of exploration, Jizhijia found that the challenges of traditional logistics are numerous: the traditional artificial supply chain has low agility and cannot respond flexibly in the face of rising and falling demand. Relying on labor will lead to labor shortages, difficulty in getting jobs in batches quickly due to difficulties in recruiting workers and high costs, and manual error-prone, low accuracy, and difficulty in handling large quantities of goods. The semi-automated rigid supply chain has huge investment in assets, high expansion costs and high risks. Rigid automated operations are slow to adapt to complex scenarios, making it difficult to adapt to rapidly changing product strategies. The most important point is that the rigid supply chain has a long construction period and slow upgrade iterations.
In this regard, Jizhijia adopts a "flexible and intelligent" strategy to deal with it. The flexible supply chain has high agility, is able to respond flexibly to changes in demand, has a small granularity and is easy to expand, and sales of promotional sales can also be quickly shipped. Can accurately deal with complex operations, digital operations to deal with massive product types and strategy iterations. The construction period is short and the upgrade speed is fast, which is convenient for enterprises to deploy flexible and intelligent.
For example, a large e-commerce retailer officially went online in August 2016 to operate Jizhijia's "goods to people" robot system, deploying 120 robots in a 6,000 square meter warehouse. It took only two months from the implementation of the entire system to the official launch of the system. After the deployment is completed, the e-commerce picking efficiency can reach 3 times the labor efficiency, and the robot warehouse manpower is saved by 70% compared with the previous one , and the effect is remarkable.
In exploring the digital transformation of the logistics industry, Jizhijia built its own robot management system RMS. This is a multi-agent scheduling and task management platform system that handles real-time path planning, traffic management, task allocation, production capacity optimization, visual status monitoring, map maintenance and sharing, multi-model collaborative scheduling, and safety in mobile robot cluster Stop waiting for the task. RMS provides support for cloud and local deployment, provides open-standard APIs and SDKs, introduces large-scale mobile robot applications for logistics companies, and provides a solid underlying guarantee. Its core advantages are:
• Large-scale scheduling: thousands of robots can be scheduled concurrently, and task execution is efficient and accurate;
• Mixed scheduling: it can support multi-navigation robots such as SLAM and QR codes to run in the same map, and multiple models cooperate;
• Cluster path planning: real-time optimization of paths, traffic congestion management and other processing mechanisms to ensure efficient path of multi-robot systems;
• Intelligent warehouse management: it can be flexibly configured according to requirements, and can be managed from multiple dimensions such as area, floor, and multi-storage.
According to IDC prediction, by 2021, 45% of mobile robot deployments will be carried out through RaaS (Robot-as-a-Service, robot as a service), which will affect multiple business departments in the organization.
Shen Mu said that RaaS is similar to IaaS, PaaS and SaaS. From a technical point of view, RaaS virtualizes resources and capabilities to achieve dynamic allocation and configuration, and can be elastically adjusted to maximize the efficiency of resource utilization. Essentially, RaaS uses both the SaaS of robot management software and the robot hardware as a resource, which is dynamically allocated on demand.
Now relying on Microsoft's Azure cloud's IaaS and PaaS services, Azure's open microservices framework and Kubernetes services, and IoT Hub's technology, Jizhijia has implemented a technical support system for the dynamic allocation and configuration of the entire software and hardware.
In the upgrading of the supply chain, RaaS can quickly realize the upgrading of the supply chain through a flexible business cooperation model:
1. RaaS reduces initial investment and reduces operating costs: flexible investment and expansion according to the needs of logistics companies, and there is no need to configure a maintenance management team.
2. High efficiency, high accuracy and quick response, alleviating labor shortage: for example, under full-load operation, the picking efficiency of Jizhijia single table is 4 to 5 times that of traditional warehouse picking efficiency.
3. High degree of flexibility: The robot intelligent warehousing service can be activated at any scale, modularly expanded, and flexibly extended; and it can be flexibly configured according to the customer's business characteristics, deployed on demand, and expand or increase or decrease shelves, robots, and workbenches at any time.
4. Fast system iteration: The enterprise can apply the latest technology without hiring researchers, and is not limited to its current technical ability to enjoy the system iteration update and the service upgrade and value-added provided by the robot supplier.
Unstructured data blowout? AI + BI punches hard!
Huike Communications (Beijing) Network Technology Co., Ltd. (hereinafter referred to as "Huike") was established in 1998, based on the information processing technology of the Chinese University of Hong Kong. For more than 20 years, Huike has used cutting-edge technology to provide an all-media information database and public relations. Media and market intelligence solutions, social media solutions, commercial big data services, etc., to help users always grasp market information, formulate effective strategies, and accelerate the speed of enterprise digital transformation.
In the CSDN "2019-2020 China Developer Survey Report", 64% of enterprises have not yet achieved intelligence (of which 14% of enterprises have no information). In this regard, Huike said that the current difficulties and challenges faced by enterprises in digitalization and big data applications are in the face of complex unstructured big data.
For example, a well-known large-scale B2C company in the world needs to better understand the feedback, portraits, brand activity performance, multi-dimensional information of competing products, and hot topics in order to better assist product strategy, marketing decisions, and customer operations , Popular trends and other unstructured data information.
However, it is not easy to truly meet these needs, facing multiple challenges such as media selection suggestions, data collection, integration, multi-dimensional AI applications, and the construction of new business analysis platforms. Huike comprehensively uses years of data integration processing experience, multi-dimensional NLP AI technology and BI platform construction and other comprehensive capabilities to help companies build an end-to-end full-link big data synthesis from an all-media data lake to application visualization. Programs to help companies and get better results.
Full media marketing big data full link solution
At present, Huike's AI big data products are mainly for MarTech and FinTech. Based on massive all-media big data, artificial intelligence technology, professional data science models, high-availability SaaS and DaaS products, it provides enterprises with multiple scenarios. Intelligent digital solutions. Behind this, Microsoft Azure to provide strong technical blessing: comprehensive coverage of technology, simple and easy to use cloud platform management and high time the data processing, to help conduct hui SaaS platform, DaaS delivery, channel partners, localized deployment Promotion of multiple business cooperation programs.
Huike currently owns a Chinese all-media database covering more than 570,000 data sources, 85 billion pieces of stock data, and more than 9.6 million pieces of data are added daily. In 2014, Wisers AI Lab was established and won multiple awards worldwide.
In addition, in order to better meet the different needs of various enterprises, Huike has created its own "Huike Branch", allowing data to be integrated into the platform in a unified and real-time manner and applied to different front-end products. Can process data in real time, millisecond-level processing delay, query data. It can also analyze multi-dimensional data and output results, using a high-performance intelligent question and answer engine to quickly display data. In terms of security, it has enterprise-level security certifications, ranging from source data, data processing, data storage, and platform security certifications.
Want to see more latest practices and applications of digital transformation?
On April 17-18 , Microsoft hosted the Microsoft Online Tech Forum with the theme of "Digital Transformation Acceleration". The above 8 big men gathered in the "Digital Transformation Best Practices" forum to discuss digital transformation together ~
The conference has set up 8 major technical forums, covering artificial intelligence, big data, AIoT, open source tools, codeless low code, cloud native, DevOps, security compliance and other cutting-edge technical issues and innovative applications. Look at the exciting topics first:
4.17 schedule
4.18 schedule
It is also revealed here that this conference is personally led by Microsoft CEO Satya Nadella !
Bring together 60+ global top technology experts and industry leaders to form a lecturer group to comprehensively interpret your dry goods experience in digital transformation:
What are you waiting for?
Immediately scan the QR code or click to read the original text
Participate for free + draw prizes + communicate with Daniel
If you want to know the details of the conference in advance, you can add a small assistant WeChat, reply to " Microsoft Conference ", and join the group for free:
Click to read the original text, participate for free and meet the future!
