< 返回


Embedded artificial intelligence (AI) is the application of machine and deep learning in software at the device level. (Image by Gordon Johnson from Pixabay)

嵌入式人工智能(AI)是机器和深度学习在设备级别的软件中的应用(图片来源:Pixabay,Gordon Johnson)

For a number of years, artificial intelligence has been at the top of the technology must-watch list for evolving trends and applications. With our ability to build smart machines that simulate human intelligence, the implications for technological advancement across numerous sectors are endless. So, what could be better than artificial intelligence? Embedded artificial intelligence.

多年来,人工智能一直处于不断发展的趋势,并位于必须关注的技术列表中的首位。随着我们越来越有能力构建模拟人类智能的智能设备,对众多领域的技术进步带来了无穷无尽的影响。那么,还有什么比人工智能更好的呢?—— 嵌入式人工智能。

Embedded artificial intelligence (AI) is the application of machine and deep learning in software at the device level. Software can be programmed to provide both predictive and reactive intelligence, based on the data that is collected and analyzed.


Over the past several years, an important shift has occurred from cloud-level to device-level processing of artificial intelligence tasks, data and results. Embedded AI is the direct result of this important shift. Traditionally, complex AI computations, such as producing search engine results, were performed at a data center in the cloud. With the implementation of AI models on graphics processing units (GPUs), session border controllers (SBCs), and systems on chips (SoCs), there is less of a dependence on the cloud for AI data processing.

过去几年,人工智能的任务、数据和结果的处理发生了从云端到设备端的重要转变。嵌入式 AI 是这一重要转变的直接结果。传统上,复杂的 AI 计算(例如生成搜索引擎结果)是在云中的数据中心执行的。随着 AI 模型在图形处理单元(GPU)、会话边界控制器(SBC)和片上系统(SoC)上的实施,AI 数据处理对云的依赖减少了。

With embedded AI, devices have the ability to run AI models at the device level and then directly use the results to perform an appropriate task or action. The cloud is still helpful from a data storage perspective, as data can be stored temporarily at the device level and eventually sent to a cloud server for safekeeping.

借助嵌入式 AI,设备能够在设备级别运行 AI 模型,然后直接使用结果来执行适当的任务或操作。从数据存储的角度来看,云仍然很有帮助,因为数据可以暂时存储在设备级别,并最终发送到云服务器进行安全保管。

While the uses and application of embedded AI are vast, here’s a short list of industries where this technology is automating processes, providing advanced analytics and business insights, and improving customer service, among numerous other benefits.

嵌入式 AI 的用途和应用非常广泛,下面是一个简短的行业列表。在这些行业中,嵌入式 AI 技术正在实现流程自动化、提供高级分析和业务洞察力、改善客户服务质量,以及许多其他好处。

  • 农业(Agriculture)
  • 航空(Aviation)
  • 现场服务管理(Field Service Management)
  • 金融(Finance)
  • 卫生保健(Healthcare)
  • 制造业(Manufacturing)
  • 零售(Retail)
  • 船运(Shipping)
  • 供应链(Supply chain)

As the technology behind embedded AI continues to evolve, two evolving applications to keep a beat on include embedding AI capabilities onto custom SoCs and Internet-connected devices or the Internet of Things (IoT). Embedding AI models on SoCs optimizes the chip’s architecture thereby reducing instruction counts, power consumption, and calculation time. While embedding AI in Internet-connected devices sounds like both a slippery slope and the next logical step, a number of companies, such as Google, Siemens and HPE, have already dipped their toes in this space.

随着嵌入式 AI 背后的技术不断发展,两个不断发展的应用领域包括:将 AI 功能嵌入到定制 SoC 和物联网(IoT)中。在 SoC 上嵌入 AI 模型可以优化芯片的架构,从而减少指令数、功耗和计算时间。虽然在联网设备中嵌入人工智能听起来既是一个滑坡,但也是下一个合乎逻辑的领域,像谷歌、西门子和 HPE 等许多公司都已经涉足这个领域了。

In manufacturing and industrial settings, pairing embedded AI and the Internet of Things can result in predictive maintenance for equipment, increased operational efficiency, improved products and services, and enhanced risk management. Of course, the applications for embedded AI and IoT devices in other settings abound, with implications in security and monitoring systems, smart homes and cities, scientific research, and healthcare to name a few.

在制造业和工业环境中,将嵌入式 AI 与物联网结合起来可以实现设备的预测性维护、提高运营效率、改进产品和服务并加强风险管理。当然,嵌入式人工智能和物联网设备在其他环境中的应用比比皆是,对安全和监控系统、智能家居和城市、科学研究和医疗保健等都有影响。


Was this article helpful?
5 out of 5 stars

1 rating

5 Stars 100%
4 Stars 0%
3 Stars 0%
2 Stars 0%
1 Stars 0%
Please Share Your Feedback
How Can We Improve This Article?