Wireless Multimedia Sensor Networks

Project Description

Wireless Multimedia Sensor Network Architecture

Traditionally, research on algorithms and protocols for sensor networks has focused on scalability, i.e., how to design solutions whose applicability would not be limited by the growing size of the network. Flat topologies may not always be suited to handle the amount of traffic generated by multimedia applications including audio and video. Likewise, the amount of processing required on each node in terms of computing and communications may result in the convergence of several research areas for the development of an efficient and flexible architecture for WMSNs.

In Fig. 1 we introduce a reference architecture for WMSNs. The figure depicts three sensor networks with different characteristics, possibly deployed in different physical locations.

The first cloud on the left depicts a single-tier network of homogeneous video sensors. A subset of the deployed sensors have higher processing capabilities, and are thus referred to as processing hubs. The union of the processing hubs constitutes a distributed processing architecture. The multimedia content gathered is relayed to a wireless gateway through a multi-hop path. The gateway is interconnected to a storage hub, that is in charge of storing multimedia content locally for subsequent retrieval. Clearly, more complex architectures for distributed storage can be implemented when allowed by the environment and the application needs, which may result in energy savings since by storing it locally the multimedia content does not need to be wirelessly relayed to remote locations. The wireless gateway is also connected to a central sink, which implements the software front end for network querying and tasking.

The second cloud represents a single-tiered clustered architecture of heterogeneous sensors (only one cluster is depicted). Video, audio, and scalar sensors relay data to a central clusterhead, which is also in charge of performing intensive multimedia processing on the data (processing hub). The clusterhead relays the gathered content to the wireless gateway and to the storage hub.

The last cloud on the right represents a multi-tiered network, with heterogeneous sensors. Each tier is in charge of a subset of the functionalities. Resource-constrained, low-power scalar sensors are in charge of performing simpler tasks, such as detecting scalar physical measurements, while resource-rich, high-power devices are responsible for more complex tasks. Data processing and storage can be performed in a distributed fashion at each different tier.

Fig. 1 - Wireless multimedia sensor network and its sub-classications.

Many challenges arise with such an architecture, that need to be solved in order to enable seamless functioning without any disruption at the user's end. As an example, consider the following considerations:

  • Sensing coverage. In traditional WSNs, sensor nodes collect information from the environment within a predefined sensing range, i.e., a roughly circular area defined by the type of sensor being used.Multimedia sensors may be sensitive to the direction of data acquisition. In particular, cameras can capture images of objects or parts of regions that are not necessarily close to the camera itself. However, the image can obviously be captured only when there is an unobstructed line-of-sight between the event and the sensor. 

  • Single-tier vs Multi-tier Sensor Deployment. A single-tier network is created when homogeneous sensors are dispersed in the region of study and each sensor is programmed to perform all possible application tasks. In the multi-tier approach, resource-constrained, low-power elements are in charge of performing simpler tasks, such as detecting scalar physical measurements, while resource-rich, high-power devices take on more complex tasks. For instance, a surveillance application can rely on low-fidelity cameras or scalar acoustic sensors to perform motion or intrusion detection, while high-fidelity cameras can be woken up on-demand for object recognition and tracking.

  • Thus, the choice of hardware, the nature of the application, the type of data sensed, the terrain of deployement are amongst the many factors that go into deciding the architecture for WMSNs and this is an active area of research.

    Below, we show some pictures of existing WMSNs.

    Cyclops, created by Henry Samueli, School of Engineering and Applied Science, UCLA, and Agilent attaches to Mica motes and uses miniaturized cameras similar to those used in cell phones, and developed the circuitry and software that enables the nodes to process the images in context and report any new information.
    http://www.engineer.ucla.edu/news/ 2006/cyclops.html


    Intel Research Pittsburgh has developed IrisNet (Internet-scale Resource-Intensive Sensor Network Services) that allows to query webcams and other sensors are spread throughout the environment through the Internet.
    http://www.cmu.edu/corporate/news/2004/ 0504extra/0504_IrisNet.html

    [Back to top]

    Research Challenges in Designing a WMSN

    The main challenges in the design of wireless multimedia sensor networks are as follows:

    • QoS requirements. Streaming media, system snapshots, audio/video store and play-back applications have different requirements with respect to delay, jitter, and loss tolerance that necessitate a fresh look at providing essential QoS services. 

    • Bandwidth. The available bandwidth is severely limited as WMSNs require transmission bandwidth that is orders of magnitude higher than that supported by currently available sensors.

    • Power. Power consumption is of greater concern than in traditional wireless sensor networks as multimedia applications produce high volumes of data, which require high transmission rates, and extensive processing. 

    • User monitoring facility. Integration with Internet (IP) architecture is of fundamental importance for the commercial development of sensor networks to provide services that allow querying the network to retrieve useful information. 

    • Leveraging of in-network support. Distributed databases of multimedia data, including distributed storage and indexing of data within the network itself, in-network processing techniques need to be developed to efficiently extract relevant information from received data.

    [Back to top]

    Overview of the Application, Transport, Network and MAC layer Multimedia Support

    Recent focus on WMSN research has resulted in advances in the understanding of energy-constrained wireless communications, and the integration of advanced multimedia processing techniques in the communication process. Another crucial issue is the development of flexible system architectures and software to allow querying the network to specify the required service (thus providing abstraction from implementation details). At the same time, it is necessary to provide the service in the most efficient way, which may be in contrast with the need for abstraction.

    We thus observe that the task of enabling efficient multimedia communication cannot be limited to the application layer rules alone, or be brought about my enhanced physical layer technologies in isolation. There is an obvious need to merge functionalities at each of the layers of the protocol stack so that multimedia transmission requirements are met. Of interst especially is a cross-layered approach that blurs the distinguishing lines between successive layers thus allowing the designer greater freedom, but at an increased software and hardware complexity. We believe that this is most suited for devising an optimal approach and begin by identifying the important networking tasks at each layer and their cross-layer dependencies, if any.

    We next analyze the factors that influence multimedia communications at each layer of the protocol stack in order to leverage the support already present and identify regions for improvement. The subsequent discussion provides an indicative example of this approach:

    • Application Layer

    The services offered by the application layer include: (i) providing traffic management functionalities (ii) providing primitives for applications to leverage collaborative, advanced in-network multimedia processing techniques; (iii) performing source coding according to application requirements and hardware constraints, by leveraging advanced multimedia encoding techniques.

    1. QoS requirements of diffrent multimedia traffic classes have can be considered as application admission criteria. Application admission control methods may determine admissions based on the added energy load, application rewards and other metrics.

    2. Of interest in WMSNs is the coding of video streams. The design objectives here are twofold: (i) High Compression Efficiency in order to effectively limit bandwidth and energy consumption, and (ii) Low-complexity encoders as they are to be embedded in sensor devices. Often these two interests are conflicting and further work needs to be undertaken to identify an acceptable solution.

    3. Different multimedia sensor network applications are extremely diverse in their requirements and in the way they interact with the components of a sensor system. Development of middleware that allows sensor use in these heterogeneous applications is an important research direction.

    • Transport Layer

    In applications involving high-rate data, the transport layer assumes special importance by providing end-to-end reliability and congestion control mechanisms.

    1. The preference of UDP over TCP may have to be re-thought in context of WMSNs as packet dropping in congestion may lead to delays in the order of a fraction of a second. This effect is even more pronounced if the packet dropped contains important original content not captured by inter-frame interpolation, like the Region of Interest (ROI) feature used in JPEG2000 or the I-frame used in MPEG2.  

    2. Despite the existence of reliable transport solutions for WSN, none of them provide real-time communication support for the applications with strict delay bounds. Therefore, new transport solutions which can also meet certain application deadlines must be researched.

    • Network Layer

    This layer addresses the challenging task of providing variable QoS guarantees depending on whether the stream carries time-independent data like configuration or initialization parameters, time-critical low rate data like presence or absence of the sensed phenomenon, high bandwidth video/audio data, etc. Some of these functionalities are described below:

    1. In the case of large WMSNs like Irisnet, it is required that the individual nodes be monitored via the Internet. Such an integration between a randomly deployed sensor bed and the established wired network becomes a difficult research challenge. 

    2. Various types of routing protocols exist, each tailormade to a particular application need. Most routing protocols that consider more than one metric, like energy, delay etc., form a cost function that is then minimized. The choice of the weights for these metrics need to be judiciously undertaken, and is often subject to dynamic network conditions.

    • MAC Layer

    As seen earlier, multimedia traffic, namely audio, video, and still images can be classified as separate service classes and subjected to different policies of buffering, scheduling and transmission.The choice of contention free (TDMA or multichannel) protocols or contention-based protocols results in different design parameters as described below:

    1. If TDMA is used as the underlying MAC layer technology, the length of the frames and the frequency of the slot reservation period can be varied while designing a multimedia system as they directly affect the per-hop latency. 

    2. Multichannel protocols results in increased bandwidth utlization which is essential for high data rate applications but this comes at a cost of greater system complexity and hardware demands.

    3. Other MAC parameters like the length of the packet, the scheduling strategy, error control and FEC mechanisms need to be investigated for best system performance.

    [Back to top]

    Investigating New Physical Layer Technologies

    We envision the coexistence of several physical layer technologies with different characteristics, including Bluetooth, Zigbee and UWB. However, with respect to the needs of WMSNs, we believe that UWB is the most promising technology due to the following reasons:

    • Particularly appealing for WMSNs are UWB high data rate with low power consumption, and its positioning capabilities. Positioning capabilities are needed in sensor networks to associate physical meaning to the information gathered by sensors. 

    • Impulse UWB or I-UWB provides spreading in frequency by transmitting pulses of very short duration. This directly translates to reduced energy expenditure, a vital criterion in the resource hungry multimedia applications. 

    • The interference caused to active devices is minimum as UWB transmission appears like noise power to the other receivers.

    Before adopting UWB as the underlying transmission structure for WMSNs, questions on efficient sharing of the medium, and providing latency and throughput bounds need to be answered. Thus, this area is still in a nascent stage and we intend to undertake further work in the course of this project that clearly defines a UWB platform for multimedia data transfer.

    [Back to top]

    Correlation-based Communication Protocol Design

    Visual contents (still images and video streams) are the dominating parts of multimedia data. Compared to scalar data, visual contents require more sophisticated techniques to process and much higher bandwidth to deliver. In a densely deployed WMSN, there exists correlation among the visual information observed by cameras with overlapped field of views. We propose to investigate the correlation characteristics of visual contents in WMSNs. More specifically, the following challenges will be addressed:

    • Determine correlated nodes.

    • Evaluate the degree of correlation.

    • Exploit correlation in both the spatial and the temporal domain.

    • Study the joint effect of multiple correlated sensors.

    • Derive a distortion function based on the above studies.

    The result of this analysis will enable us to design multimedia in-network processing techniques. For example, according to specific application requirements, each sensor node can filter out uninteresting events locally, or coordinate with each other to aggregate correlated data. Furthermore, correlation-based communication protocols can be developed. The functions of the lower layers, such as admission control, routing decisions and congestion control, can be designed to be aware of the correlation characteristics. By making use of the correlation characteristics, we hope to achieve efficient and reliable multimedia communication in WMSNs.

    [Back to top]

    You are visitor:

    since 8/16/2006.