Embedded reasoning is seeing a surge of interest driven by advances in embedded computing power and the desire to control increasingly complex systems safely, efficiently, and reliably. It incorporates the strengths of AI reasoning — planning, scheduling, controlling, learning, and diagnosing — into physical systems. This advances system capabilities in solving complex tasks, in acting on high-level goals, and in adapting to changing and uncertain states. Applications are being revolutionized, from robotics to transportation systems to industrial automation. The integrated methods by which we approach these problems are also rapidly evolving. As illustrated by recent research programs and applications, these emerging capabilities require a tight integration of diverse techniques with a strong multi-disciplinary understanding of their relationship. The traditional interfaces of the fields of AI reasoning, control, and human factors are becoming blurred, with control system optimization running on embedded processors, and artificial intelligence controlling vehicles. To enable systems of sensors, actuators, and processors to be adaptive, distributed, and robust, many challenges must be addressed. Software execution must occur in real-time with process and communication concurrency to enable interaction with the physical world. Systems must understand their environment and act intelligently and often autonomously, with potentially noisy sensor inputs and imperfect models. They often integrate capabilities such as inference, strategic and tactical planning, optimal behavior selection, reactive control, and fault tolerance. Systems must interact appropriately with each other and transparently with their human operators. New programming paradigms are evolving to address these challenges. Embedded reasoning systems draw upon a multitude of technologies.