publication computing edge
  • Published on 22 Feb 2021
  • About ARTEMIS

Embedded Artificial Intelligence: The ARTEMIS Vision

“Advances in embedded and cyberphysical systems have disrupted numerous application domains. We examine the requirements and challenges of these technologies, which present significant opportunities for interdisciplinary research.” Thus Dimitrios Serpanos, who wrote a column on Embedded Artificial Intelligence on behalf of the ARTEMIS Scientific Council for IEEE Computer and the Computing Edge.

As he continues:

Embedded computing has brought significant advances in application domains ranging from home appliances and health systems to environmental monitoring and from smart factories to autonomous transportation (cars, trains, ships, and airplanes) and smart cities. Embedded computing systems constitute the cyber part of cyberphysical systems (CPSs). Autonomous CPSs are commonly used in processes of increasing complexity that are designed and implemented with single-processor systems (for example, a patient’s insulin pump) or distributed, interconnected processing nodes (for example, autonomous vehicles). Autonomous CPSs have also become increasingly connected to the Internet of Things (IoT), which includes specialized networks, such as the Industrial IoT, Internet of Vehicles, and others. Clearly, a hierarchy of CPSs is emerging, where simple autonomous systems are interconnected to create higher-level autonomous systems that, in turn, are interconnected to provide even more complex systems and applications. For example, a CPS for an autonomous car’s cruise control is part of an autonomous car—a more complex distributed CPS—that may be a node of a network of autonomous vehicles (a fleet) managed through a cloud application. The pervasiveness of embedded systems and the increasing deployment of CPSs lead to an emerging infrastructure that spans globally and enables the development of new applications and services that were infeasible or inconceivable in the recent past. The immediate availability of operational data as well as computational power in conjunction with artificial intelligence (AI) techniques provides significant opportunities for systems and services worldwide. To achieve this vision, CPSs need to be efficient, scalable, and extensible in terms of both hardware and software. The adoption of CPSs in various application domains leads to strong constraints on their design and implementation. More precisely, CPS technologies are quite demanding for the purpose of satisfying strong application and operational environment requirements, including real-time constraints, safety and security, continuous operation, scalability, extensibility, autonomy, power consumption, and internetworking. Although CPSs typically abide by several of these requirements, the application domains, ranging from manufacturing to transport and from health to power, impose different constraints on each specification. For example, industrial production systems have stricter requirements for real-time constraints and continuous operation than home automation systems, while they have more relaxed stipulations for power consumption relative to autonomous, mobile health-monitoring systems.

Download the Computing Edge issue of January 2021 here. The full article is featured at page 8.