The vision of future networking is that not only people but also all things, services, and media will be connected and integrated, creating an Internet of Everything (IoE). Internet-of-Things (IoT) systems aim to connect and scale billions of devices in various domains such as transportation, industry, smart home/city, medical services, and energy systems. Different wireless and wired technologies link sensors and systems together, through wireless access points, gateways, and routers that in turn connect to the web and cloud-based intelligence. IoT architectures make great demands on network control methods for the efficient management of massive amounts of nodes and data. Therefore, some of the cloud's management tasks should be distributed around the edges of networked systems, utilizing fog computing to control and manage, e.g., network resources, quality, traffic prioritizations, and security. In this work, we present adaptive edge computing solutions based on regressive admission control (REAC) and fuzzy weighted queueing (FWQ) that monitor and react to network quality-of-service (QoS) changes within heterogeneous networks, and in a vehicular use case scenario utilizing IEEE 802.11p technology. These adaptive solutions are providing more stable network performance and optimizing the network path and resources.
Operators, developers and manufacturers are striving to become part of the Internet of Things (IoT) and Internet of Everything (IoE) revolution, creating new types of products and systems. IoT systems with connected devices and things will cover the whole world and affect all people globally. Hence, networked intelligence will spread to various application domain areas including industry, Intelligent Transportation Systems (ITS), wearables, health, smart homes, offices, buildings, grids and cities. Typical IoT architecture can be categorized broadly into four interconnected systems including things, gateways/routers, networks and clouds. Efficient IoT architecture requires that the things and sensors must be intelligent enough to filter and manage the data that they send to the cloud. However, many of the current sensors were initially not designed to be connected to the Internet and are not capable of processing and sending data to the cloud, although there can be great amounts of data flowing around. For example, a jet engine may produce 10 TeraBytes of data about its performance and conditions in only 30 minutes of flight, according to Cisco . To transfer all the data into a cloud and the response back to the system without any pre-processing would consume not only the scarce bandwidth resources but the time and money of different players in the IoT product chain. In response to this problem, part of the network intelligence and data management should be distributed to gateways and routers in the interconnected systems creating fog  and edge computing operations.
This paper proposed adaptive computing methods for IoT networking at the network edges to optimize and control traffic flows and network resources. The fog computing challenges at the edge routers includes e.g. QoS issues, network provisioning and resource management. With the REAC method, the adaptive edge router monitors the link performance to admit the flows to the network in a way which handles congestion and preserves good quality for prioritized users. The QoS scheduling capabilities utilize FWQ to control traffic flows according to the prevailing traffic level in a smooth and fast way in heterogeneous networks.
The developed mechanisms are able to react faster to traffic changes and guarantee better quality for prioritized traffic and at the same time preserving fairness to other flows than the traditional control and scheduling methods without adaptive characteristics. The developed overall system reacts to changes in the network QoS by determining decision making procedures on the possible flow rejection, marking, or allowed bandwidth weight assignment, thus bringing cognition to the network path. In future work, the adaptive traffic management methods need to be evaluated and the scalability tested in a large-scale environment for combining the different algorithms optimizing the performance of the IoT applications. Testing these features as SDN and NFV components would also be beneficial for the resource usage optimization.