Search

International audience; Modern vehicles are equipped with several interconnectedsensors on board for monitoring and diagnosispurposes; their availability is a main driver for the developmentof novel applications in the smart vehicle domain. Inthis paper, we propose a Docker-based container platform asa virtualization solution to implement customized smart carapplications. Through a proof-of-concept prototype—developedon a Raspberry Pi3 board—we show that a container-basedvirtualization approach is not only viable but also effective andflexible to manage several parallel processes running on board.More specifically, the platform can take priority-based decisionsby handling multiple inputs, e.g., data from the CANbus basedon the OBD II codes, video from the on-board webcam, andso on. Results are promising for the development of in-vehiclevirtualization techniques in future vehicles.

The development of small form-factor handheld electronics is pacing the personal devices market, followed by the increasing number of various applications. Some of those applications also cover computation-hungry use-cases, such as image or video processing and compression, among others. Historically, wearable and handheld devices were not designed to execute computationally intensive operations for reasons ranging from limited battery capacity to radiated heat. Offloading computationally heavy tasks to a comparatively more powerful and less energy-dependent device can help prolong the battery lifetime of a wearable. This paper analyzes different task offloading scenarios from the wearable to a device located at the network edge. Such a device can be a smartphone paired with the wearable or an edge server co-located with the cellular base station. A comprehensive performance evaluation conducted under a wide variety of realistic settings in terms of task demands, processing capabilities, and data rate, unveils the circumstances in which offloading is convenient and when it is not, in terms of meaningful metrics. acceptedVersion

In this paper the authors' attention is on the specification of a middleware functionality for dynamic QoS control in multimedia cellular environments, such as UMTS. The functionality implements a soft QoS paradigm, at the same time optimising the perceptual QoS provided to the end user according to his/her needs. It will be shown that by exploiting suitable performance indexes and resource redistribution algorithms both an acceptable grade of service and a good degree of user satisfaction can be achieved, without an excessive increase in the network-signalling load.

In this paper a dynamic QoS management technique for an enhanced UMTS environment is proposed. It aims at continuously matching user preferences and resource constraints imposed by the mobile environment. A point of strength of the proposed mechanism is represented by the user profiling procedure adopted. It foresees that the user has associated a set of preferences and requests which are automatically taken into account by the resource allocation and negotiation algorithm. This allows to contemporary maintain a high level of user satisfaction, and contemporary, to reduce the network signalling load.

We analyze a dynamic channel assignment mechanism which selects the full-duplex channel that provides the minimum sum of signal strength measures on both uplink and downlink channels. In a previous work such a mechanism has been shown able to provide a remarkably better reuse degree than the "optimal" fixed channel assignment algorithm. In this paper we improve the analysis considering the impact of call admission control policies on system efficiency and flexibility. We show that a call admission algorithm based on the same signal strength measures used for the selection slightly impairs the reuse efficiency, but allow the system to exploit traffic fluctuations.