Sensor fusion for smart containers

The transport of goods around the globe is a major commercial activity, vital to the world´s economies. Much of this is accomplished using large standard sized containers which can be readily on and off loaded from different platforms for travel by road, rail, air and sea. Criminals and terrorists may target such containers for their own purposes. The threats include using the containers to assist with smuggling (drugs, currency or arms), stowaways (illegal immigrants or terrorists) and terrorist attacks (weapons of mass destruction). At present, the checking of containers is limited, and hence they are vulnerable prime targets. The various means which criminals may use in order to achieve their objectives include delays, accidental or deliberate damage, theft of goods or services, fraud, piracy, and electronic attacks. In an effort to combat these threats and improve security, smart containers are being introduced. These contain a variety of sensors to detect, identify, log and communicate what happens to a container and its contents during its journeys around the world. The types of sensor may include: door opening sensors; volume intrusion alarms; passive infra-red sensors; microwave radar; ultra wide band radar; sound, light and temperature sensors; ultrasound sensors; visible light and infra-red cameras; gas sensors (e.g. carbon dioxide); chemical, biological, radiological, nuclear and explosive sensors; electronic tag readers; depth gauges; GPS receivers and inertial sensors. In addition to the sensor data from a container, other relevant information must be taken into account, for example the expected location of the container, and its mode of transport. This will involve acquiring and utilising data from other sources, for example, the bill of lading, traffic reports and weather information. The particular problem addressed in this paper is the processing of the sensor data and other information to determine whether or not a container is under attack, and to communicate this to the authorities in a timely manner. Of major importance is that the probability of detecting a genuine attack must high, and of falsely indicating an attack must be low. Moreover, t- he cost of implementation must be kept low for both the initial installation of the system and the servicing of it during operation. In the paper we will report on our ongoing investigations to design and test an algorithm to fuse the data from the various sources in order to decide whether or not an attack has been made, and to provide the operators with justification for the decision.