Internet of Things

The Internet of Things (IoT) presents an enormous potential in transforming the way we live and do business. With an estimated 500 billion IoT devices in the next decade, the IoT provides a global connectivity of sensors and actuators to form a computer network of unprecedented nature and with large challenges to support connectivity of this massive number of devices as well as designing scalable network and storage systems able to service these new demands.

Research trends in IoT have focused heavily on technologies to guarantee that communication across devices takes place properly and have the potential to scale well at a local level. Although this is definitely an important focus, our interest goes beyond the mechanisms to collect data at a local level. This is driven by the fact that the new wave of Internet deployments requires mobility support and geo-distribution in addition to location awareness and low latency between communicating peers and to Cloud infrastructure for further analysis and storage. This demand for flexibility and responsiveness leads to the introduction of novel deployment models such as fog computing as a natural evolution of cloud computing, which is accompanying IoT in today’s widespread sensor and actuator network deployments. Fog computing provides a hierarchical and distributed platform that scales well with a massive number of data sources and exploiting heterogeneous devices, from servers to sensors.

Furthermore, researchers have found it useful to move away from the Internet’s point-to-point communication abstraction and instead adopt constructs that are more data-centric. Data-centric applications extend across basically all application domains such as energy, agriculture, manufacturing, logistics, telecommunications, space and IoT and big data are coined terms that often are introduced concurrently when digitization is addressed more broadly. 

However, many challenging issues still need to be addressed and a plethora of technological knots need to be loosened before the IoT vision becomes a reality. Connectivity to the “things” is typically provided via low-power wireless radios deployed in harsh and heterogeneous environments on computing platforms that are constrained with respect to processing, memory capabilities, and the available power for the corresponding devices. Network deployments are often so large that self-configuration and self-healing properties are required. Consequently, research in IoT enabling technologies must deal with the efficient use of communication channels, cost-effective transmit and receive processing schemes and computer resources as well as methods and tools for efficient operation of large networks. Finally, the development of a cohesive infrastructure that is able to support the various requirements of different IoT applications and that is designed jointly with Big Data processing algorithms is crucial to the future and effectiveness of IoT services. 

Our research activities concerns IoT as enabling technology in digitalisation. More specifically, our research activities will focus on the following four areas: 

  1. IoT connectivity
    The IoT relies on the underlying paradigm of Machine-to-Machine (M2M) communications to integrate a plethora of various sensors, actuators, and smart meters across a wide spectrum of businesses. Today, the M2M landscape features diversity of available connectivity solutions, Ultra-Reliable and Low-Latency Communications (URLLC), and massive machine-type communications. Use cases of URLLC include autonomous vehicles, monitoring and control, tactile feedback, control and coordination of unmanned aviation vehicles, robotics, and industrial automation. "Internet of things" will address the URLLC research challenge including relevant use cases linking to the industry.
    The IoT connectivity challenge furthermore embraces the need for innovative digital transmission schemes, error control coding with high performance and low latency, enhancing adaptive protocols to ensure robust and reliable communications on constrained computing platforms. This includes the design of novel modulation approaches, medium access control schemes and light-weight protocols. These efforts also link with activities in cyber security protocols and encryption schemes from WP3 as the resulting protocols require a joint design.
  2. Robustness and signal reconstruction
    Digitalisation has permitted the creation of sensing and processing systems that are more robust, flexible, and cheaper than their analogue equivalents. Sparse or compressible representations of communication signals have fuelled research in compressive sensing and sparse signal processing with the promise of offering signal reconstruction with high fidelity. Compressive sensing and sparse signal processing enables a potentially large reduction in the sampling and computation costs for sensing signals that have a sparse or compressible representation and is developing to become an invaluable asset in IoT networks. Moreover, low-resolution signal processing techniques also has the potential equip low-power IoT devices and offer energy-efficient solutions.  Signal characteristics in IoT networks include correlation in the data that could be in the time domain (e.g., measurements from the same device) or in the spatial domain (e.g., measurements from multiple sensors). Compressive sensing can provide a means to reduce overall traffic to the Cloud, while still been able to recover the most relevant information from a set of sensors (decoding requires less samples than the total samples in the system). This form of compression is also very attractive to IoT as computation in the sensors is minimal (e.g., sample and transmit) and can be carried out in a distributed fashion (i.e., compression across multiple sensors without coordination). Although the decoder requires computationally demanding algorithms, these can be run using the vast resources in the Cloud. This asymmetry of computation is beneficial for IoT, particularly, because standard compression algorithms typically have a more complex encoder for compression and a simpler decoder (e.g., video, audio), which would put the burden in the sensors. Moreover, the ability to compress across multiple sources is not typical in compression algorithms and is advantageous for applications that have a large spatial correlation, e.g., temperature/humidity in different points of a farm will be quite similar at the same time of the day.
    Our key challenge is to develop compressive sensing for the IoT and understand the fundamental trade-offs between quality of measurements and resource-usage, including, computation, network traffic, and energy at the IoT devices. We will also explore its interplay with techniques such as network coding in order to further reduce the costs on the IoT devices and network use. The use of compressive sensing together with other techniques can have a direct impact on the IoT connectivity requirements (e.g., the system can tolerate nodes being offline or disconnected), offered load by the devices (e.g., less transmissions per device reduces the congestion in the network), and energy used by the devices (e.g., sensors may be powered off for longer periods because their information can be reconstructed from other nearby sensors and/or the sensors future/past transmissions).
  3. Enabling technologies for IoT (fog/cloud storage and computing)
    Although Fog computing reduces latency and provides mobility support, it also imposes a new set of challenges for the transmission and storage of information, which will curb its potential in practice. These include (a) introducing a large number of copies of the same content across the network, i.e., higher costs for storage, where a device or user will only be able to access one at a time, and (b) making data consistency complex across the system, because there will be different versions of the content (e.g., from dynamic data in IoT as identified by EUDAT in different locations across the network, requiring a large amount of network traffic to enforce consistency and signalling to coordinate data access. The latter is particularly critical in the presence of mobile devices, which may move across the network and use different resources during their transit, e.g., in the Internet of moving things. Beyond local storage and caching, Fog computing does not provide more efficient storage mechanisms to reduce the storage costs in the Cloud, which now needs to cope with the massive amount of data generated by IoT. The Cloud typically stores data reliably by storing multiple copies of each file (data replication), which incurs in large economical costs. Our key focus for storage in the Cloud is on theory and systems that can reduce the storage space used such as erasure codes and network coding, but also focusing on novel approaches that combine these erasure codes with massive-scale compressions techniques, e.g., data deduplication. Efficient storage techniques are not only critical to reduce the amount of space used for storage of data, but they are fundamental to allow algorithms for Big Data processing to efficiently operate on data that are naturally stored in multiple devices across the network. For this reason, this area of research will maintain strong interactions with WP4’s Big Data analysis algorithms developed on top of our storage solutions.
  4. Internet of moving things
    A novel set of Internet applications that combine IoT with a high degree of mobility among network nodes is emerging. This calls for the design and development of systems and methods for supporting a network of mobile nodes. Typically, scenarios involve robotics and machinery working in a collaborative effort or loosely connected networks such as networks of autonomous vehicles and drones, low-earth-orbiting satellite network constellations. Current communication networks fall short to adequately support communication environments involving mobile and static nodes. This often results in inefficient routing and loss of connectivity in the network. Experiments with using Long-Range Radio Radio (LoRa) gateways for IoT devices will be carried out to explore the optimal routings here.