Mobile Agent-based Data Management Framework for Wireless Sensor Networks

0 Introduction

With the rapid development and maturity of low-power radio communication technology, embedded computing technology, satellite sensor technology and integrated circuit technology, it is possible for a large number of low-cost satellite sensors to self-organize into wireless sensor networks through wireless links. Wireless sensor network has the characteristics of self-organization, self-adaptability and fault-tolerance, and has a wide application prospects in many important fields such as military, industrial and agricultural, biomedical, environmental monitoring and so on.

Wireless sensor network consists of a large number of sensor nodes and a few sink nodes. Sensor nodes located in the monitoring area are responsible for collecting related data information and eventually transmitting the data to the sink node. Wireless sensor networks are data-centric and aim to collect data with some characteristic attributes, regardless of which sensor node the data comes from. Based on these ideas, database researchers Abstract wireless sensor networks as data systems, separating the physical implementation of wireless sensor networks from their applications [2]. Because the data characteristics of wireless sensor network are large amount of data, strong timeliness, network nodes have limitations in energy, calculation, storage and communication capabilities [3], good data management technology can improve the efficiency of sensor nodes cooperative perception, collection, processing, and publishing data information, thereby better determining the availability and effectiveness of wireless sensor network.

Mobile Agent is a distributed computing technology with many advantages. This paper presents a framework based on mobile agent technology to improve the existing data management mode of wireless sensor network. Using mobile agent not only collects data but also can process data, reduces the load on data management center, and overcomes the limitations of current data management scheme.

1 Mobile Agent Mechanism [4]

Mobile Agent is the product of the combination of distributed computing technology and artificial intelligence technology. It is essentially a program entity with certain intelligence and judgment. It can migrate between network nodes under its own control and according to certain rules, find and process appropriate resources, and complete specific tasks on behalf of users. When an agent is created in one execution environment, it is able to migrate its state and code across the network to another environment to resume execution.

The characteristics of mobile agent are mainly reflected in the following aspects:

A) Autonomy: Once an agent is initialized, it can execute independently. Without subsequent direct intervention, it can control its own internal state and external behavior, and it can also be authorized to make certain decisions and complete related tasks.
B) Reactivity: Agents can perceive and act on their environment and respond to changes in the environment, which can be the physical world, users using graphical interfaces, other collections of Agents, or a combination of all these.
C) Motivation: Agents not only respond to their environment, but also actively exhibit goal-oriented behavior.
D) Communicativity: The interaction between agents adjusts to the task-driven changes and collaborates with each other to complete complex tasks.
E) Mobility: Agents can migrate from one node to another at any time and autonomously on a network. The running Agent state can be stored and transferred to the target node, and suspended Agent programs can be resumed for execution.

2 Data Management Architecture

At present, data management of wireless sensor network mostly uses the similar client/server data processing mode [5], which can be divided into two types: centralized data management structure and distributed data management structure.

The centralized data management structure is characterized by relatively independent processing of perceived data and access to wireless sensor networks. Sensed data of each sensor node is routed to a centralized center or server after multiple hops. The data management decision of the network is made by the processing center. This method is simple, but because all sensor data needs to be transferred to the central processing node, the communication cost is large and the scope of use is small.

Distributed data management structure is also called intranet processing structure. Considering the large-scale distribution of wireless sensor network nodes, the possible energy consumption loss and communication overhead caused by a large amount of redundant information transmission, the distributed data relationship structure is characterized by that some sensor nodes in the network are responsible for some data management tasks, such as performing aggregation operations, in addition to data forwarding tasks.

3 Data Management Framework Based on Mobile Agent

In the real data service environment that supports sensor networks, the above architectures have deficiencies in network bandwidth, load balancing, and dynamic adjustment of data processing strategies due to node failures.
The idea of data management based on mobile agent is to use Agent to host the data in the network on the nodes that get it, to propagate and process the code and status of specific information needs through Agent or computing engine in the network, and to return the final information to the data management center (converged node or central server).
A mobile agent-based wireless sensor network is shown in Figure 1.

Combining the distributed characteristics of wireless sensor network data and the assignment of data management tasks, the network data management framework based on mobile agent mainly includes two parts: mobile agent entity and mobile agent service environment. As shown in Figure 2.

The mobile agent service environment is part of the system, which provides a transparent, easy to control, secure and reliable data management execution environment. It is mainly responsible for the creation, operation, suspension, termination, transmission and reception of mobile agent. The service environment determines the migration and collaboration strategies of agents and the transmission and communication mechanisms based on the pre-set initial task execution scenarios.

Mobile Agent entities, including service agent and messenger agent, are the dispatchers and executors of network data management tasks. Service Agent resides in mobile Agent service environment, providing access support and task decision for Messenger Agent. Service Agent sends mobile agent programs according to MACP (Mobile Agent Control Protocol), and controls its behavior based on network data status information. Messenger Agent moves between network data sources, using control over resources to satisfy the purpose of remote data source access.
Compared with the current data management architecture processed in the network, the mobile agent-based data management framework is based on mobile agent technology, which hosts the agent support environment on node devices, the data management center based on the mobile agent service environment, sends agent programs with appropriate functions according to the actual network state to target nodes, generates program instances in the Agent running environment, and accomplishes the expected functions. This management framework essentially uses Agent's flexible data collection and computing power to move processing functions to data sources locally, so that the computing limits required for data management are distributed throughout the network. Sensory data of each sensor node is managed independently by Agent, and the performance situation can be acquired in time. In addition, the negotiation and knowledge learning mechanisms between Agents can be used to dynamically implement decision-making on data processing tasks. Satisfies sensor network performance optimization.

4 Internal structure of mobile agent

Mobile Agent for network data management refers to a program that is dynamically published to the data source node and executed at the target node according to the requirements of data management. The purpose of this program is to optimize management. Agent behavior (start, hang, terminate, delete) is totally controlled by the process of sending it. As shown in Figure 3, the structure of mobile agent includes communication layer, inference machine, management layer and Agent knowledge base. The internal behavior of a mobile agent is that the communication layer receives information from the external environment (messages and requests from other agents), and the inference engine (including a priori set of rules related to data processing) interactively learns from the knowledge base based on these messages, requests, and the state of the agent.

The relational task decision-making layer of the service agent makes a reasonable task judgment based on the result of the inference engine. The task is assigned to the messenger agent through the communication layer. After the messenger agent gets the assignment task, it performs the migration action. After reaching the target node, the messenger agent performs the related action (collection, fusion, query, etc.) after analyzing and inferring the known sensor node data, and updates the Agent status information. And determine the next set of administrative tasks to complete. When the Messenger Agent completes the task, the result is sent to the Service Agent.

5 Communication mechanism between mobile agents

The mobile agent communication layer supports the negotiation process in distributed decision making and receives execution request, notification or result information from other agents. Negotiation requires a timely understanding of the data processing capabilities accomplished by the sensor. Communication messages between mobile agents come from the decision-making process within the agent, including identification, type, content, and sender and receiver related to data management. The priority of system data management decision task is determined by service agent, and the priority of task execution is determined by messenger agent. Performance indicators for management tasks are determined by negotiation between mobile agents.

6 Concluding remarks

At present, wireless sensor networks have many problems, such as large amount of sensing data, low wireless communication bandwidth, limited energy allocated to each network node, unreliable environment, and so on. The application of mobile agent in wireless sensor network data management will help to process a large amount of data in unstable and dynamic environment with limited bandwidth and energy, reduce development and application costs, and reduce the complexity of wireless sensor network data management to a certain extent. In the future research, the formalized representation technology of data management architecture and energy consumption analysis will be further studied.