本文来源:https://www.eetimes.com/silicon-labs-adds-iot-developer-platform-tools/
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Silicon Labs 增添 IoT 开发者平台工具
为满足不断增长的边缘 AI 需求,Silicon Labs 将其开发工具与生成式 AI 和企业数据基础设施相连接。
Silicon Labs 今日(10 月 7 日)宣布了一系列举措,让开发者更轻松地构建互联 IoT 硬件并使用边缘 AI。
在其年度Works With 2026 峰会上,这家总部位于得克萨斯州奥斯汀的公司推出了软件开发工具、一个开源软件项目以及一项云集成合作。这些旨在帮助工程师编写代码、检查硬件设计,并在低功耗微控制器上部署机器学习模型。
这些发布之际,芯片厂商正面临企业客户要求简化嵌入式软件开发的压力。有限的内存、算力和射频选项常常拖慢产品上市进度。Silicon Labs 表示,公司不仅仅提升芯片性能,还在通过将开发者工具与生成式 AI 助手和企业数据平台相连接,来应对软件开发挑战。
面向嵌入式软件的 AI 工具
本次更新的关键一环是Simplicity AI SDK的公测发布。Silicon Labs 表示,该软件套件让开发者以有序的方式访问其文档、工具和代码库,供 GitHub Copilot 和 OpenAI 的 Codex 等 AI 编程助手使用。首个版本面向低功耗蓝牙(BLE)项目,让 AI 智能体协助完成配置、构建、烧录、调试和网络分析。
上周,在正式发布之前,Silicon Labs 产品经理 Tamas Daranyi 将该工具描述为提升效率的手段,而不是严谨工程的替代品。
“它不是解决方案,也不是魔法,但它是一个工具、是非常严肃的赋能,”Daranyi 在布达佩斯的 Pretzl Connect 发布活动上告诉 EE Times,并指出开发者可以使用该 SDK 加速产品定制和概念验证测试。
Daranyi 回应了对使用生成式 AI 代码质量的担忧,强调开发过程中人的监督仍然必不可少。“你仍然是开车的人,”他告诉我们。“现在快多了,但它仍然是一辆车,而开车的是你。”他还解释说,工程师或许不必阅读 AI 生成代码的每一行,但仍需彻底测试代码,确保它在硬件上能够运行。
连接硬件意图与代码
除 SDK 之外,Silicon Labs 还推出了Simplicity Design Intelligence。这一软件框架帮助把高层级的产品构想转化为可运行的嵌入式配置。其首个功能名为 Hardware Intent,它会查看硬件规格和电路板图,帮助在实际制板之前选择引脚和外设。
在布达佩斯的简报会上,Silicon Labs 公关负责人 Sam Ponedal 表示,硬件设计需求通常以许多不同格式出现,包括技术 PDF、手写笔记或视频讲解。Hardware Intent 会接收这些自然语言输入,并把它们转换成软件智能体可用的格式。
“Hardware Intent 可以接收所有这些输入——自然语言,无论什么格式——并把它们转换成扁平化的、智能体就绪的输出,开发者可以立即用智能体来构建应用,”Ponedal 说。他解释说,该系统能够自动发现细微问题,例如原始项目需求与电路板图之间引脚数量不匹配或外设分配缺失。这有助于避免昂贵的硬件重新设计。Silicon Labs 计划于 2027 年 1 月发布 Hardware Intent 的 Alpha 版本。
开源 BLE 软件
Silicon Labs 表示,公司还在面向开源开发者社区启动公测,首先从其低功耗蓝牙(BLE)软件协议栈开始。开发者可以在 GitHub 上访问示例应用和工具、提交问题、建议代码修复并贡献拉取请求。
当开发者提出修复建议时,Silicon Labs 的工程团队会自动加人工地进行检查。“如果它满足那些测试需求,我们就会把它实现进我们的 SDK,使其成为 Simplicity SDK 发布版本的正式组成部分,”Ponedal 在布达佩斯告诉我们。
Daranyi 解释说,该开源项目覆盖应用层软件和开发工具,但不包括更底层的射频协议。“我们不对协议中的所有内容开源,这一点非常重要,”Daranyi 说。专有的射频层和底层协议细节将保持闭源,而更高层级的示例应用将开放,以鼓励社区协作。
边缘端的企业级 MLOps
为帮助直接在微控制器上运行 AI,Silicon Labs 与数据管理公司Databricks达成了合作。这项合作将边缘设备连接到 Databricks Data + AI 平台,使工程师能够在其常规企业工作流中完成数据收集、模型训练和嵌入式优化。
首个Silicon Labs MLOps SDK让联网边缘设备能够将遥测数据发送到 Databricks 进行模型训练。训练完成后,Silicon Labs 表示其转换工具可以把 PyTorch 或 ONNX 模型转换为 TensorFlow Lite。
Databricks 内置的基于 Python 的 ML Profiler 会检查模型在目标硬件上的运行情况,在部署前提供周期数、内存使用和能耗等信息。
这项集成解决了一个常见的边缘 AI 问题:数据科学团队通常在云端工作,而嵌入式工程师使用专用微控制器。通过把性能分析和优化工具直接加入 Databricks,开发者无需单独的环境即可改进边缘模型。
另见:
Pablo 是一位拥有 30 多年经验的资深工程师。10 多年来,他一直担任 EE Times 特约编辑(现为供应链栏目编辑)。他还为 EPSNews、InformationWeek、EBN、LightReading、Network Computing 和 IEEE Xplore 撰稿。他的报道涵盖供应链、半导体、网络、物联网、安全和智慧城市。他拥有俄亥俄州立大学电气与电子工程硕士学位。 在 LinkedIn 上关注 Pablo
Source: https://www.eetimes.com/silicon-labs-adds-iot-developer-platform-tools/
For copyright issues, please email: 13951431913#139.com
Silicon Labs Adds IoT Developer Platform Tools
To meet rising edge AI demands, Silicon Labs connects its development tools with generative AI and enterprise data infrastructure.
Silicon Labs today (Oct. 7) announced initiatives to make it easier for developers to build connected IoT hardware and use edge AI.
At its annual Works With 2026 Summit, the Austin, Texas-based company introduced software development tools, an open-source software project, and a cloud integration partnership. These are meant to help engineers write code, check hardware designs, and deploy machine learning models on low-power microcontrollers.
These announcements come as chipmakers face pressure from enterprise customers to make embedded software development easier. Limited memory, computing power, and radio options often slow down product launches. Rather than improving only silicon performance, Silicon Labs said it is working to address software development challenges by linking developer tools with generative AI assistants and enterprise data platforms.
AI tools for embedded software
A key part of the update is the public beta release of the Simplicity AI SDK. Silicon Labs said this software kit gives developers organized access to its documentation, tools, and code libraries for use with AI coding assistants such as GitHub Copilot and OpenAI’s Codex. The first release targets Bluetooth Low Energy projects, letting AI agents help with setup, building, flashing, debugging, and network analysis.
Last week, ahead of the formal announcement, Tamas Daranyi, product manager at Silicon Labs, described the tool as something that boosts efficiency, not as a replacement for careful engineering.
“It’s not the solution, it’s not magic, but it’s a tool and very serious enablement,” Daranyi told EE Times at the Pretzl Connect press event in Budapest, noting that developers can use the SDK to accelerate product customization and proof-of-concept testing.
Daranyi addressed worries about code quality when using generative AI, stressing that human oversight is still essential during development. “You are still driving the car,” he told us. “Now it’s much faster, but it’s still a car, and you are the one driving it.” He also explained that engineers might not have to read every line of AI-generated code, but they still need to test the code thoroughly to make sure it works on the hardware.
Bridging hardware intent and code
Alongside the SDK, Silicon Labs introduced Simplicity Design Intelligence. This software framework helps turn high-level product ideas into working embedded setups. Its first feature, called Hardware Intent, looks at hardware specs and board diagrams to help choose pins and peripherals before building the actual boards.
At the Budapest briefing, Sam Ponedal, head of PR at Silicon Labs, said that hardware design requirements often come in many different formats, including technical PDFs, handwritten notes, or video walkthroughs. Hardware Intent takes these natural-language inputs and turns them into a format that software agents can use.
“Hardware Intent can take all of those inputs, natural language, no matter what format, and it can turn them into a flattened, agent-ready output that developers can immediately start using an agent to build their application,” Ponedal said. He explained that the system can automatically spot subtle problems, such as mismatched pin counts or missing peripheral assignments, between the original project requirements and the board diagrams. This helps avoid expensive hardware redesigns. Silicon Labs plans to release an alpha version of Hardware Intent in January 2027.
Opening up BLE software
Silicon Labs said it is also starting a public beta for its open-source developer community, beginning with its Bluetooth Low Energy (BLE) software stack. Developers can access sample applications and tools on GitHub, submit issues, suggest code fixes, and contribute pull requests.
When developers suggest fixes, Silicon Labs’ engineering team will check them both automatically and manually. “If it meets those testing needs, we’ll implement it into our SDK, so it becomes an official part of the Simplicity SDK release,” Ponedal told us in Budapest.
Daranyi explained that the open-source project covers application-level software and development tools, but not the lower-level radio protocols. “It’s very important that we don’t open source everything in the protocol,” Daranyi said. Proprietary radio layers and lower-level protocol details will stay closed, while higher-level sample applications will be open to encourage community collaboration.
Enterprise MLOps at the edge
To help run AI directly on microcontrollers, Silicon Labs has partnered with Databricks, a data management company. This partnership connects edge devices to the Databricks Data + AI platform, so engineers can handle data collection, model training, and embedded optimization as part of their regular enterprise workflows.
The first Silicon Labs MLOps SDK lets connected edge devices send telemetry data to Databricks for model training. After training, Silicon Labs said its conversion tools can convert models from PyTorch or ONNX into TensorFlow Lite.
A Python-based ML Profiler inside Databricks checks how models will run on the target hardware. It gives information such as cycle counts, memory use, and energy consumption before deployment.
This integration solves a common edge AI problem: Data science teams usually work in the cloud, while embedded engineers use specialized microcontrollers. By adding profiling and optimization tools directly to Databricks, developers can improve edge models without needing a separate setup.
See also:
Ambient IoT: From Battery-Free Promise to Mass-Market Reality
Smarter Buildings, Safer Occupants: Intelligence Meets Privacy
Pablo is a seasoned engineer with 30+ years of experience. For over 10 years, he's been a contributing editor for EE Times (now editor of the Supply Chain section). He also wrote for EPSNews, InformationWeek, EBN, LightReading, Network Computing, and IEEE Xplore. His coverage spans Supply Chain, Semiconductors, Networks, IoT, Security, and Smart Cities. He holds an MEng, Electrical and Electronics Engineering from The Ohio State University. Follow Pablo on LinkedIn
留下回复取消回复
你必须注册或登录后才能发表评论。