Description
Date depot: 1 janvier 1900
Titre: Networks Flexible and Energy Efficient Wireless Sensor for Critical Infrastructures Surveillance
Directeur de thèse:
Vania CONAN (Thales)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
Context and motivations
Recent developments in the field of telecommunications and electronics have put online objects that, traditionally, were not allowed to communicate with the surrounding world, driving to the creation of an Internet of Things [5]. This, potentially, brings to scenarios where billions of interconnected devices become parts of a whole big network since they become addressable by remote entities. The scenarios where these “smart” objects operate are often constrained, in the sense that, for instance, battery replacement may be costly or simply impossible. Even when recharging is possible, most applications require low-power devices and protocols to reach acceptable lifetimes. Moreover, communication links quality changes frequently due to fading effects or the mobility of the devices, and the storage and processing capabilities of such devices are really limited. Therefore, the optimization of resources becomes a central issue.
To enhance current systems for protection of critical infrastructures, existing traditional, high-power and bulky-sensing systems (e.g., radar, optronic) are going to be supported by the deployment of low-power sensor networks, since they offer a cost-effective complement or alternative to more. Miniaturized devices are often used to monitor specific and localized phenomena (e.g., vibration, intrusion, chemical). When networked with each other, they offer a high fidelity image of situations with a fine granularity and a low rate of false positives. The smart miniaturized objects involved in critical infrastructure surveillance are calling for flexibility of operation and deployment (e.g., plug’n’play, re-purposing), distributed intelligent processing, energy-efficiency, and robust secured data transportation. Howerer, whilst traditional protection systems requires high efficiency in terms of throughput and delay, but really low energy efficiency, sensor networks have exactly the opposite requirements, therefore an adaptation strategy, focus of this thesis, is required.
Objectives
Caching has been introduced in metering applications to reduce the number of transmissions when several requests ask for the same information [7, 9, 10]. However, it can be also extended to other application needs and used for other types of optimizations. Defining what kind and how information should be collected and how they should be exploited is an important research challenge. Indeed, beyond usual data flows, information such as network statuses (topology, service availability, energy levels, subscriptions) can be exploited to optimize the network configuration with regards to external demands or to reduce redundant signaling mechanisms inside the protocol stacks. However, unless caching is done cleverly, this mechanism easily increases latency, as the packets wait at intermediate nodes, or could provoke inconsistency when retaining control packets that report status in a highly dynamic environment. The caching system should ideally treat data and control packets differently, possibly taking inspiration from the QoS mechanisms that classify traffic into classes (DiffServ, 802.11e, etc.). This thesis will extend the usage of cache in sensor networks by proposing a caching architecture and mechanisms to exploit this information, with the target of minimizing energy consumption and maximizing system lifetime.
As hardware is evolving fast and provide continuously new interesting capabilities, we will apply caching exploitation strategies to the management of emerging hardware features. Surveillance systems [11] are typically deployed for long-lasting operations on batteries and must react quickly in case of an event or potentially vehicle multimedia streams such as images or small videos. To enable this flexibility, we will consider new hardware platforms such as OPAL [8] which integrate several radio technologies that can be activated and deactivated when needed. This thesis will then work on mechanisms that allows to switch from a low-power to a high-throughput and low-latency network. Caching information of data and control information will be used to make such decisions. In this multi-radio context, the nodes will be able to activate the high-throughput radio only when needed, using the low-power radio to activate it. Behind this intuitive description, this scenario introduces research challenges that will constitute the starting point for this thesis. The diverse radio interfaces may not have the same characteristics. For example, low-power radio is generally either limited in terms or range (e.g. Ultra Wide Band, Bluetooth LE) or in terms of throughput (e.g. proprietary technologies in the 433MHz band). The topologies induced by the two technologies may not be identical and waking up a node may require multi-hop communication. Bandwidth limitation may introduce collisions and frame losses, which impose to include explicit reliability mechanisms in the
Doctorant.e: Leone Remy