A Proactive Complex Event Processing Method for Large-Scale Transportation Internet of Things

The Internet of Things (IoT) provides a new way to improve the transportation system. The key issue is how to process the numerous events generated by IoT. In this paper, a proactive complex event processing method is proposed for large-scale transportation IoT. Based on a multilayered adaptive dynamic Bayesian model, a Bayesian network structure learning algorithm using search-and-score is proposed to support accurate predictive analytics. A parallel Markov decision processes model is designed to support proactive event processing. State partitioning and mean field based approximation are used to support large-scale application. The experimental evaluations show that this method can support proactive complex event processing well in large-scale transportation Internet of Things.

The Internet of Things (IoT) bridges the gap between the physical world and its representation within the digital world. In recent years, with the rapid development of information and communication technologies, bandwidth and storage are no longer restricted in IoT applications. The key issue is how to process the massive events produced by large-scale IoT applications (an event means an atomic occurrence of interest in time).

In IoT applications, event processing engines need to process events that arrive from various kinds of sources such as sensors, RFID readers, and GPS. The events generated by devices directly are called primitive events. The semantic information inside primitive events is quite limited. In real application, users pay more attention to high level information such as business logic and rules. For example, each reading operation of the RFID reader at a garage generates a primitive event, but complex events like “the car leaves the garage” are the kind of events that users are really concerned with. To get complex events, many primitive events need to be combined based on some rules. In IoT application systems, business logic is converted into complex events and business logic is processed based on complex events detection. Complex event processing (CEP) [1] is used to process massive primitive events and get valuable high level information from them. As an example, in logistics industry, CEP is used to track the goods and trigger some actions when exception is found.

In many real-time IoT applications, events are uncertain due to some factors such as measuring inaccuracy, signal disturbance, or privacy protection. Usually, we use probabilities to process such uncertainty. Therefore, it is necessary to develop probabilistic event processing engine.

The traditional type of the event processing is reactive processing which means that actions are triggered by events or by the system states. A proactive event processing system is able to mitigate or eliminate undesired future events or states or to identify and take advantage of future opportunities, by using prediction and automated decision making methods [2]. For example, in a transportation system, we can predict some possible congestion states and take some actions to mitigate or eliminate the congestion states. The proactive event processing systems use predictive analytics (PA) technology which predicts future events or system states through analysis of historical events. The system also uses iterated decision processes (DP) technology which analyzes system states and selects appropriate actions to achieve expected states. The CEP, PA, and DP technologies have been studied widely, but there are few papers about how to integrate them together to support proactive event-driven system. Proactive event processing in large-scale transportation IoT needs to process massive historical events and analyze complex states iteratively which makes most of the existing algorithms unable to be used directly. Furthermore, proactive event processing systems need high performance since they usually have to take actions in time.

In this paper, we propose a proactive complex event processing (Pro-CEP) method for large-scale transportation Internet of Things. We designed a multilayered adaptive dynamic Bayesian network (mADBN) model for predictive analytics. Based on probabilistic complex event processing, our method uses concurrent actions Markov decision processes (MDP) to integrate CEP and PA.  Continue reading

DCL:  This is a technical paper dealing with the prediction of events that are yet to happen in an IoT system in order to more efficiently filter and process events that will be of interest.  Whether the ideas proposed here work or not, I do not know.  But if they do work, they will become very important  in the future.