A survey of multimedia streaming in wireless sensor networks

Satyajayant Misra, Martin Reisslein, Guoliang Xue
2008 IEEE Communications Surveys and Tutorials  
A wireless sensor network is comprised of small, low powered self-organizing sensor nodes, densely deployed in the area to be monitored. These networks now support a wide range of applications, such as health monitoring and surveillance of a military battlefield. Some of these applications may be augmented by the use of real-time multimedia. Real-time multimedia applications have stringent requirements for end-to-end delay and loss during network transport. These requirements can be categorized
more » ... on the basis of the layer of the network stack at which they arise. In this survey, we classify the mechanisms that have been proposed for multimedia streaming in wireless sensor networks at each layer of the stack. Specifically, we consider the mechanisms operating at the application, network, and MAC layers. We also review existing cross-layer approaches and propose a few possible cross-layer solutions that combine the best approaches at each of the layers, to optimize the performance of a given wireless sensor network for multimedia streaming applications. . I. INTRODUCTION Recent advances in embedded systems and wireless communications have led to the creation of wireless sensor networks (WSNs), consisting of low-cost, low-power, multi-functional sensor nodes (SNs), that are small in size and communicate over short distances [6] . These tiny sensors have sensing, data processing and communication components and are able to communicate wirelessly over multiple hops with the help of their neighboring SNs. They can be used for continuous sensing, event detection, location sensing, and local control of actuators. These nodes run algorithms to self-organize into a network and communicate among themselves and the BS(s). A WSN is composed of a large number of SNs, deployed densely and most times randomly in the area being monitored. In general, the SNs in a WSN sense data and convey them to one or more high power nodes called the sink or the base station (BS) which do most of the complex processing. The sink (or BS) might be the final destination of the data or might act as a hub from where the data is sent to users over the wired network. Improved versions of the SNs have the ability to perform local processing computations, sending only the relevant part of the sensed data. This is an important improvement over their predecessors as it has been shown that the SNs typically spend most of their energy in transmitting data to the sink (or neighbor node in case of multi-hop scenario). Thus, in-house processing often results in a reduction in the overall energy consumption. Data sent at a periodic rate from a source across the network to a destination is commonly referred to as a data stream [30] . Data streaming is an inherent feature of a WSNs because most sensors send data across the network at a periodic rate towards the BSs. Sensors continue sending data so long as they detect the entity they are observing or there is a requirement for data upstream. With the wireless video sensor nodes coming of age, a new albeit nascent field of research has sprung up, namely exploration of the possibility of use of video/acoustic sensors to set up WSNs, to monitor and convey data in the form of only video, only audio, or both video and audio streams. Video and audio streams are often referred to as real-time data streams. These streams are similar to the data stream with two important differences, they are larger in size and require various guarantees for bandwidth, delay and jitter from the network. Streaming of real-time data in the WSNs expands the realm of usability of the currently available SNs. Real-time data in the form of sound and/or video, along with the other sensed information that the WSNs already provide, have the ability to augment the field of remote surveillance, structural monitoring, and a host of other WSN applications. 3 The ability of WSNs to provide support for video/audio streaming is restricted due to the hardware, bandwidth, and power limitations of the SNs that make-up the network. Hardware is limited because the currently available cameras for WSNs cannot be made small in size and image obtained from them are of low quality having limited use. Even with good image quality, multimedia communication inherently requires high bandwidth. A WSN is labored to provide enough end-to-end bandwidth for such communications. In addition, SNs are power-starved as they run on small batteries that have limited power. In such a scenario, the goals of being able to send multimedia data streams over the network and maintain/extend the lifetime of the sensors become two diametrically opposite requirements. This implies that there is a need to define the requirements of a real-time multimedia communication in a WSN keeping the abilities of the network in mind. The requirements for multimedia streaming need to be tuned to the properties of the WSN to ensure that the nodes in the network survive longer and provide high quality sensed data. A few applications, such as [13] and [49], have been proposed for multimedia streaming in WSNs. These applications address some of the issues that need attention for supporting multimedia streaming in a WSN environment. In most of these applications video/audio sensors are used in conjunction with other sensors, such as motion, acoustic, heat, and light sensors. These other sensors do the initial sensing and identify and locate the target. Once the target is located, the video/audio sensors can be used to send periodic high quality data updates of the target. This mechanism helps in conserving the energy of the SNs while also improving network lifetime. The motivation for this survey is to provide an insight into the basic requirements for real-time multimedia streaming in WSNs and how they have been satisfied. We also highlight the issues that we believe need to be resolved. We approach the issues on a per-layer basis, associating each issue with the layer of the network stack at which it arises. We classify the solutions proposed at each layer based on the techniques used, giving their salient features and their merits and demerits. In addition, we also indicate open issues at each layer. The rest of this survey is structured as follows: In Section II we describe the characteristics of the WSN and the SNs that make up the network. In Section III, we present the quality of service (QoS) requirements for multimedia streaming applications in WSNs. Section IV highlights the requirements of a multimedia streaming application (MSA) from the application layer perspective, we classify the approaches proposed and analyze them. In Section V, we highlight the requirements of an MSA from the 4 network layer perspective and survey the QoS solutions at the network layer. In Section VI, we present the requirements of an MSA from the MAC layer and we review the schemes. Section VII discusses cross-layer optimization approaches to solve the issues for supporting MSAs in a WSN. We also propose a few solutions based on the cross-layer optimization approach. In Section VIII, we conclude our survey and outline possible directions of research. II. CHARACTERISTICS OF WIRELESS SENSOR NETWORKS Even though the WSNs are generally considered to belong to the family of wireless mobile ad hoc networks they have many characteristics that differentiate them from mobile ad hoc wireless networks. WSNs are more resource constrained, deployed in a much larger scale, and are data centric. In general, they consist of low power, multi-functional SNs (e.g., TelosB motes [36]), powered by small irreplaceable batteries. These SNs are densely deployed and are capable of sensing and conveying data towards the BS(s). The BS (sink) in a WSN is assumed to have power and memory that are orders of magnitude more than that of the SNs. The wireless medium in use is characterized by high path loss, channel fading, interference, noise disturbances, and bit error rate (BER), these result in the wireless channel having much lesser capacity than wired channels [24] . The throughput of a wireless channel fluctuates significantly owing to its timevarying characteristics, making long-term bandwidth predictions difficult. In addition, the SNs may be mobile to various extents. For instance, in some WSNs the sink or the clusterheads may be mobile [51] . Consequently, the topology in a WSN is highly dynamic, resulting in frequent changes to the routes. The small transmission power of the SNs necessitates multi-hop communication to reach the BS. In fact the path loss and channel fading characteristic of the wireless medium make multi-hop communication economical [6] . However, multi-hop communication tends to generate more interference, delay, and both higher packet loss and error during transmission. Interference and high packet loss rate on the path affect the bandwidth and delay values of the route. WSNs have a data centric nature, the data delivery model in use defines the energy requirements to a great extent. The data delivery models in WSNs could be categorized as continuous, event driven, query driven, or hybrid [44] . In the continuous model each SN transmits data periodically. In the event and query driven model the data transmission is triggered by an event or a query respectively from the BS. The hybrid model is a combination of both continuous and event or query driven model. The MSAs conform Optimize Utility given Constraints Input− multimedia (content characteristics, required QoS, etc.) Station/Node Constraints (delay, power, etc.) System Constraints (fairness, etc.) Output Different Layers Parameters (degree of adaptability can be limited)
doi:10.1109/surv.2008.080404 fatcat:g634lgxe3va4bgow4wx46y6uqi