Mobile multimedia sensor networks: architecture and routing

Recent advances in the fields of wireless technology, multimedia communications [1] and intelligent systems [2] have exhibited a strong potential and tendency on improving human life in every facet, including entertainment, socialization, education, and healthcare. To enable smart multimedia services in a mobile and ubiquitous environment, video surveillance system [3] may interface with other technologies, such as wireless sensor networks (WSNs), wireless multimedia sensor networks (WMSNs) [1, 4, 5], body area networks [6], Human computer interaction [7], intelligent agent system [8‐11], and (cooperative) multi-antenna communication networks [12‐15]. With hardware advances, this paper investigates the employment of some powerful sensor node, which is equipped with both mobility and multimedia functionalities, and proposes Mobile Multimedia Sensor Networks (MMSNs). When controlled by contextual information collected by other systems, MMSNs can further support interactive and mobile multimedia services. In this case, the marketing opportunities for advanced consumer electronics and services will expand, and more autonomous and intelligent applications will be generated. Yet, various research issues regarding node mobility, coverage, and multimedia streaming over mobile environments are still in clouds for MMSNs.

In this paper, we first present the architecture of MMSNs. Then, we focus on multimedia delivery with the strict quality of service (QoS) requirements. By utilizing location information, we design a routing algorithm with QoS provisioning in an energy-efficient manner. The routing algorithm is called mobile multimedia geographic routing (MGR), which are designed to minimize energy consumption and satisfy constraints on the average end-to-end delay of specific applications while constructing multiple paths to the sink node along the moving trajectory. MGR has the inherent scaling property of geographic routing, where packet-delivery decisions are locally made, and the state at a node is independent of the number of nodes in the network. Most importantly, it achieves flexible energy-delay trade-offs.

Notation used in this paper is given in Table 1. The rest of the paper is organized as follows. Section 2 presents related work. The architecture of MMSNs is presented in Section 3. Section 4 gives an illustrative application for MMSNs. We describe the MGR scheme in Section 5. Simulation model and experiment results are presented in Section 6. Section 7 concludes the paper.

Since the proposed MGR is geographic routing scheme for QoS provisioning in mobile multimedia sensor networks, we will introduce the related work in three aspects, i.e., wireless multimedia sensor networks, geographic routing, QoS provisioning for delay sensitive traffic in WSNs.

Due to node mobility in MMSNs, some multimedia sensor nodes can move to various critical locations for collecting comprehensive information such as image or video stream. Previously, the issue of guaranteeing soft QoS delay for delivering multimedia streams while prolonging lifetime over a bandwidth-limited and unreliable sensor network is addressed by exploiting multiple node-disjointed paths, in order to achieve load-balancing, reduction of path interference, enlarged bandwidth aggregation and fast packet delivery. However, those work are targeted at multimedia transmission over static WSNs [1]. In comparison, the proposed MMSNs have the following features:

Figure 1 shows a simple illustrative architecture of MMSN. When a mobile multimedia sensor node (MMN) moves in MMSNs, it periodically sends a multimedia flow at a new location. If a geographic routing scheme is used, the MMN sets up an individual path to the sink node for each multimedia flow. As time goes on, a series of paths will be built up while the MMN moves along a certain trajectory. Given the illustrative scenario shown in Figure 1, the sequence of the constructed paths to transmit multimedia traffic to the sink could be: Path-A, Path-B, Path-C, Path-D, Path-E. If the mobility mode and multimedia collections are controlled by other systems intelligently, more and more automated applications can be generated for industry and daily life.

Figure 2 is an illustrative application of enabling location-aware mobile multimedia services for healthcare. In this application, Tom is an old person and needs care. He owns a smart house, where three RFID readers are deployed at the proximity of the three entrances to his house. A certain number of sensor nodes are deployed in his house to detect environmental parameters. In addition, there is a mobile multimedia sensor node, which is equipped with camera, as shown in Figure 2. To save energy, mobile node is powered off if no tasks are detected.

Once Tom enters his house through one of the entrances, his ID information stored in his tag will be transmitted to the nearby RFID reader. With the awareness of Tom's ID, the mobile node is activated. The system will periodically collect the three RSSI (received signal strength indication) values from the three RFID readers (i.e., RFID reader 1, RFID reader 1, and RFID reader 3 in Figure 2) to estimate the location of Tom. Assume the result of RFID-locating is living room, the mobile node will move to living room to take video for Tom. Due to his requirements for patient care, every details of his activities in specific rooms (i.e., living room and study room) need to be video-recorded. The video streaming is forwarded to the local video server through the access point. These video images are time-stamped and stored in a directory associated with Tom's profile. When Tom moves from living room to study room, the result of RFID-locating will be changed to study room, and thus the mobile node will follow Tom to study room.

In this system, the vital signals of Tom are collected by body sensors. These body signals are subsequently updated into the database through sensor node(s) and/or access point. Any abnormalities that do not require immediate treatment may be logged into the database and registered by Tom's RFID tag for future reference. Based on these body signals, a diagnosis might indicate more complicated multimedia information is needed to further ensure the accuracy of the diagnosis. On the other hand, the resolution of the camera in the mobile node can be adaptively adjusted by the severity of diagnosis result according to the contextual information (e.g., Tom's profile, behaviors, body signals, and environmental parameters, etc.). It might be possible for the doctor to remotely diagnose Tom immediately through the real-time video communications through the mobile node, as well as the physiological data information retrieved by a wireless body area network hosted by Tom. It's critical for the mobile node to approach to the object in a timely and energy-efficient fashion.

Since our design goal is to effectively support the multimedia service in MMSNs, we consider the performance in terms of both delay and energy. First, the delay guaranteeing is treated as the goal with top priority for the QoS provisioning. Then, the energy consumption should be minimized to enlarge the life time of sensors. This motivates to exploit the energy-delay trade-offs for the design of mobile multimedia geographic routing (MGR) scheme.

We implement our protocols and perform simulations using OPNET Modeler [28]. The network with 2,000 nodes is randomly deployed over a 2,000 m × 1,000 m field. We let the sink node stay at a corner of the field and one MMN be located at the other corner. When simulation starts, the MMN will move back and forth along the diagonal line of the network field. We assume the sink node and the ordinary sensor nodes are stationary. Our sensor node implementation has a four-layer protocol structure. The sensor application module consists of a constant-bit-rate source, which generates delay sensitive multimedia traffic with a certain QoS requirements. We use IEEE 802.11 DCF as the underlying MAC, and the maximum radio transmission range (R_max) is set to 60 m.

We mainly consider the following four performance metrics:

First, the proposed MGR scheme is compared to a pure shortest path based routing scheme (i.e., GRSR[19]). Figure 4 shows the snapshots of two OPNET simulations. The snapshots are for the path constructions when MMN moves along the diagonal line in the scenarios of GPSR and MGR, respectively. By comparison, the paths constructed by MGR are more straight. As shown in Figure 5a, the path lengths in GPSR and MGR are similar. However, MGR's hop distance is adaptively adjusted to save energy while keeping the end-to-end objective delay. Thus, the hop count in MGR is larger than pure distance-based routing scheme, as shown in Figure 5b.

The delay requirement T_QoS is set to 0.035 s. As show in Figure 6a, both GPSR and MGR guarantee the QoS delay in most cases. In GPSR, paths have various delays ranging from 0.014 to 0.035 s. By comparison, most of the delays in MGR change from 0.025 to 0.035 s. The delay fluctuation of GPSR is much larger than MGR. It is because the GPSR does not have delay control mechanism without the consideration of MMN's up-to-dated location when it moves in the network.

As shown in Figure 6b, the energy consumption of GPSR is higher than that of MGR. It is because the maximum transmission range is always used by a greedy approach in GPSR. By comparison, in MGR, the end-to-end delay is softly guaranteed while the energy is still saved. Figure 6c shows the comparison of average energy consumption. MGR saves about 30% energy consumption when compared to GPSR. In our experiments, the simulation time corresponding to the last data point is also equivalent to the lifetime. As shown in Figure 6, the lifetimes of GPSR and MGR are 675 and 1,130 s, respectively, and MGR yields 455 s more lifetime than GPSR.

In this paper, we propose mobile multimedia sensor networks (MMSNs) where mobile multimedia sensor node (MMN) is exploited to enhance the sensor network's capability for event description. Then, the trade-offs of end-to-end delay and energy consumption for supporting multimedia service with delay QoS requirement are discussed. By utilizing location information, we design a routing algorithm named mobile multimedia geographic routing (MGR) for QoS provisioning in MMSNs. When MMN moves in the network, MGR is designed to minimize energy consumption and satisfy constraints on the average end-to-end delay of specific applications. The experiment results show the efficiency of MGR in satisfying QoS requirement while saving energy. In future, we will further improve MGR for more reliable and efficient QoS-oriented transmission scheme and adapt MGR for the scenarios with multiple multimedia flows per source-sink pair.