Each year $12.5 trillion of merchandise is traded worldwide using more than 20 million intermodal containers; 90% of these shipments are between seaports. As has been reported and documented in numerous studies and observations, such unsecured freight represents a global security threat in terms of both potential for lost or damaged merchandise and the crippling of the global trading economy. Within the more immediate (or tangible) realm of homeland security, such containerized freight could be used to transport harmful biological, chemical, and radioactive materials into the United States and its allied countries. The implications of a security "event" involving a cargo container are enormous. A Brookings Institute study estimated the GDP impact of a shipment, via container, of weapons of mass destruction to a major port: "This will cause extended shutdown in deliveries, physical destruction and lost production in contaminated areas, massive loss of life, and medical treatment of survivors. Potential cost: up to $1 trillion." Add to this the apparently endless variety of goods packed into the containers, and the parameters that sensors could measure seems almost boundless.
The reality is that a cargo container is nothing more than a large metal box with bolts, locks, and doors. Many companies have been working with various governmental agencies to try to define a single technical and logistical solution for the cargo container sensing problem. As corporate entities get involved, disagreements arise over which is the best sensing technology or wireless implementation to use. Yet within the realm of RF engineering lie tangible problems requiring realistic technical answers, the most fundamental being, "Can I put a radio into the container and still have it communicate reliably with the outside world?" It is this specific question that these presented measurements address.
First, What Frequency?
As with any wireless communication system, we must address the matter of which frequency to use and whether to operate in the licensed or unlicensed band. Atmospheric attenuation indicates that the system should operate at lower frequencies (see Figure 1). However, the situation is clouded by worldwide frequency allocations. Given that the cargo containers may be used throughout the world, for worldwide regulatory compliance, operation at (nominally) 2.4 GHz is chosen.
Figure 1. This is a comparison of attenuation levels presented to three common RF frequencies.
Implementation Implications and Network Topologies
While various networking topologies may be used in configuring systems of wireless sensors, a mesh networking architecture conveys certain key advantages for our particular application: wireless sensor systems installed inside containers. While the containers themselves are relatively low-cost metal boxes (~$1500), any installed "solution" must be capable of operating (including both sensing and data transmission) for years while running on a simple battery power source. While advances in low-power RF and sensor design are reported seemingly daily, the reality is that the RF section of the module typically consumes a few milliamperes of current while in the transmitting mode and microamperes to milliamperes of current while in various sleep states. The level of power consumption substantially increases when the RF transmit power is increased.
If the wireless sensor unit installed within a cargo container is to communicate with the outside world, the network topology used for system deployment is critical. For example, if the ever-present star (hub and spoke) topology exemplified by most 802.11 (Wi-Fi) networks is used in this application, each wireless sensor module must be able to communicate directly with the base station. Radio link performance for each node is characterized by the point-to-point radio communication between node and base station. Classic communication theory (and reality) dictates that if the attenuation present in the node-base station channel increases—due to an increased separation distance or some container cargo that absorbs more of the RF signal—system performance will degrade unless transmitted power is increased. However, increasing the radiated power consumes more of the battery, thereby decreasing the installed system's lifetime. It is a vicious circle.
In contrast, in a mesh network where each wireless node relays the information packet(s) from its nearest neighbors, the system operation isn't as dominated by the node-base link performance. There are subtle and not-so-subtle performance details that arise in implementing a mesh. Each individual node in individual cargo containers doesn't have to be in direct contact with a base station—with the accompanying higher radiated power and power consumption—it only needs to communicate with its neighbors. There must be enough mesh-networked wireless sensor nodes within contact to guarantee overall network stability and data transmission. Please understand that you don't get something for nothing, since there are system performance issues related to mesh networks, too. For a realistic assessment of performance metrics for a mesh-networked wireless sensor system, see our companion article, "The Realities of Dealing with Mesh Networks," Sensors, June 2004, www.sensorsmag.com/articles/0604/14/main.shtml.
This cargo container network situation is typified by a mesh network. The information bounces around the mesh but at some point must be extracted and presented to the outside world. Following the nomenclature of the ZigBee Alliance (ZigBee is essentially a protocol that runs on top of the wireless channel's physical layer), the information extraction process falls to the Network Coordinator. The sources of the information, i.e., the wireless sensor node transceivers, are referred to as End Devices, and the Routers serve as repeaters. If each end device can communicate directly with the coordinator, the "mesh" reverts to the classic star topology.
Now that we've set the stage, let's look at performing system operational measurements for mesh-networked wireless sensor nodes installed in cargo containers. The physical situation is depicted in Figure 2 which shows a cargo container storage area located in the Bay Area of northern California.
Figure 2. These are stacked cargo containers located in a storage yard near Oakland, CA. Notice that the height, width, length, and manner of corrugation may vary. For the test and measurement results presented here, identical 40 ft. length containers were used.
The measurement procedure involves describing the network's performance based on the packet transmission rate (as opposed to the bit error rate) for a node that is moved around within certain adjacent containers. Due to logistics in the proposed container sensing implementation, a node is attached to a locking bar on one end of the container. Another node is attached to a PC that is running the packet error rate determination software. For these measurements, the separation between the locking-bar node and the measurement-PC node is 20 m with a clear line of sight.
Figure 3. Nodes within each container were placed in positions 1–6.
Another node is placed at specific locations within the container to which the locking-bar node is attached (see Figure 3) or is placed at similar physical locations within a container next to the one with the locking-bar node (see Figure 4).
Figure 4. Only a single RF node was located outside container A, mounted to one of its locking bars. The nodes located within the containers communicated with this node, which would then relay the message to the node located at the measurement station. In certain instances the nodes within the containers could also communicate directly with the measurement station s node.
The process is as follows: The node is placed at the appropriate container position, and the packet error rate is measured and recorded. The node is then moved to the next location, and the measurement and recording process is repeated.
In the case of nodes placed within a single container with an inner node relaying information to the node on the outside front locking bar, the data communication path performance is determined by how reliably the nodes within the container can communicate with the locking bar node. Results for this scenario are shown in Figure 5.
Figure 5. The success with which information packets are relayed from nodes placed at different locations within a container to a node located on the outside locking bar of the same container is shown.
In the second case, the nodes within an adjacent container (call it container B) must relay their signals to the node mounted on container A's outside locking bar. Measurements were made for this scenario with the container B nodes shifted among positions 1–6, and these results are shown in Figure 6.
Figure 6. Packet transmission from a node placed within a container (B) to the node located on the outside locking bar of an adjacent container (A) is shown.
For nodes located within containers A and B, measurements of packet transmission rates are much more problematic. The actual transmission rate values are calculated on the measurement PC, which has a node in communication with the locking bar node. In trying to ascertain how a node located at site B2 (position 2 within container B) and a node at A3 performed, the nodes may have an operating channel directly to the locking bar node. The beauty of the mesh topology is revealed, for the packet will attempt both paths, with the measurement PC recording the values for the best one. Therefore we used a slightly different step for direct intercontainer packet transmission: The measurement PC was placed within one container, while nodes were located at various positions within the adjacent container. The results are shown in Figure 7.
Figure 7. Intercontainer packet transmission is shown. As opposed to the prior measurements, in this case the measurement PC and its associated node were placed within container A (at location 5), while another node was placed at locations 1–6 within container B.
Mesh-networked wireless sensor systems promise to provide users with an easy way to deploy nodes in various physical environments. There are significant channel and logistical problems presented by radios broadcasting information out of a metal cargo container. One method to increase the probability of successful transmission, without increasing actual transmitted power with its associated increased power consumption, is to rely on the mesh to relay the signal to the destination. Regardless of the situation, the scenario begins with getting a reasonable level of packet transmission out of the container, a situation that may be optimized through the node placement within the container. Measurements have been performed for nodes placed at different locations within a container and relaying the message to other nodes. The nodes used are 802.15.4 compliant and operate in the 2.4 GHz region. While physics suggests a lower frequency range be used, the 2.4 GHz band is (essentially) available worldwide.
As the results show, successful establishment of a reliable wireless mesh network dictates overall system performance. The issues associated with node power management lead to a tradeoff. You can either place multiple RF nodes within each container (allowing for lower radiated power and the accompanying lower power consumption, but carrying the penalty of extra cost and installation/maintenance complexity because there are more nodes present), or you can have a higher radiated power per node. Answers to these issues lie outside the realm of wireless network engineering. What our measurements show is that the optimal placement, at least for the many different ISO-compliant containers used, is to attach nodes to the container's ceiling, centered (side to side), and ~11% (of the container's length) from each end.