System autonomy is built on effective AI integration. The increasing demand for AI and ML is driving the development of accessible, efficient and specialized SDKs—often in markets where expertise is scarce and turnaround must be quick to remain competitive. Take, for example, agricultural markets. Combines, skid steer loaders and implements are now competing with features such as real-time crop monitoring, thermal imaging and wildlife avoidance. Each application demands the development of high-powered, highly specialized perception software. Developers quickly become bogged down if there is a steep learning curve to an SDK lacking a powerful toolchain, sample applications and comprehensive demonstrations. On top of this, embedded AI applications are often developed by small teams faced with the challenge of deciding between outsourcing development, adopting open-source software, or a combination of the two. To keep pace with their markets, embedded software engineers need support for a variety of OS or RTOS choices, such as Linux or FreeRTOS. As a result, engineers purchase software licenses from middleware vendors to leverage expertise and reuse code. SDKs, therefore, must support high-quality operating systems with the flexibility to incorporate third-party software stacks.

Figure 2-2 Embedded system software enables features such as real-time crop monitoring and autonomous operation.

In the humanoid robotics industry, innovation is taking place so rapidly that there is no time to wait for new product releases to begin development. The software development experience becomes a make-or-break factor for hardware adoption and, ultimately, a successful product. Simple onboarding, debugging tools and DevOps practices are major considerations. DevOps is a set of principles, including CI/CD (continuous integration/continuous delivery), that are expected by embedded software engineers from SDK suppliers to keep up with the pace of product development in such a rapidly paced industry. CI/CD is a process in which a software provider delivers updates to the product immediately, without waiting for grouped releases. In this way, developers in need of bug fixes or new features experience faster turnaround and can stay on schedule.