1 Related work Delay tolerant network DTN (delay tolerant network) is used to transmit data in intermittently connected networks. Literature [5] gives the overall structure of DTN. As an overlay network on the transmission layer during network interconnection, it provides data storage and Services such as retransmission and authenticated data transmission. DTN is based on message exchange, and its data units may be messages, packets or bundles. Burleigh et al. also proposed a new end-to-end overlay Network protocol [6], called bundling.
DTN technology has been introduced into wireless sensor networks in recent years. According to the different mobility of nodes, related research work can be roughly divided into three categories: (1) Sensor nodes are all stationary. The first type of delay tolerant wireless sensor network DTSN (delay tolerant sensor networks) are static, and all sensor nodes and convergence points are fixed. Based on the limitations of limited transmission distance of sensor nodes and low battery energy, each node is loosely connected together, and sensor nodes leave the network and become isolated nodes. This phenomenon occurs from time to time. Document [7] is an example of a specific application of static DTSN. The article proposes to use a sensor network to monitor the human living environment and gives users the authority to control this sensor network through the Internet network. Document [8] mentioned The purpose of the SENDT project is to establish a proof-of-concept sensor network to monitor lake water quality. Since some sensor nodes need to be shut down regularly to save battery energy, a discontinuously connected DTSN structure is formed between each node. .
(2) There are some easy-to-manage mobile nodes in the network. In the second type of DTSN, some mobile nodes can be used to improve the connectivity of the network. For example, the data mule method [9] is used to aggregate each sensor in a network with a sparse number of sensors. Data information collected by sensor nodes. There is a data mule entity in the network that moves randomly in the sensor deployment area. It regularly collects data collected by each node and provides interactive information storage and forwarding services. Due to the large transmission distance required by sensor nodes Therefore, this strategy can save the energy of the sensor more than the static DTSN.
(3) The sensor node is moving, that is, DTMSN. The most critical thing is how to design the routing algorithm in the DTMSN environment. The most basic routing algorithm (data collection method) is called direct transfer [10]. The basic idea is that the sensor node Communication only occurs with the sink point, that is, only when the sensor node moves into the communication range of the sink point, the two will transmit messages. Obviously, the transmission energy consumption of this strategy is very low (because the sensor node only communicates with the sink point), And the message transmission delay mainly depends on the communication frequency between each sensor node and the sink point: the lower the communication frequency, the greater the message transmission delay. Since the communication frequency between sensor nodes and sink points in DTMSN is often low, the The transmission delay of the algorithm is large and the data transmission success rate is also low. In response to the above problems, literature [11] proposed a flooding algorithm, which allows the sensor node to copy the message to all nodes within its communication range, with the purpose of making all nodes in the network Each node contains a copy of this message. The size of the node storage queue and the packet loss strategy after the queue is full have a great impact on the performance of this algorithm. If the storage queue is large enough, this algorithm can achieve the desired performance at the expense of consuming a large amount of sensor energy. Higher data transmission success rate. However, due to the limited storage queue of sensor nodes, this algorithm causes serious packet loss and poor performance. In addition, ZebraNet [12] uses mobile sensors to monitor the living habits of zebras. He based on historical records Routing: Each sensor node saves the historical level of its successful forwarding of data packets directly to the convergence point. When a sensor node meets another sensor node, only when the historical level of the former is less than the latter, the two will generate data. transmission. However, this simple strategy does not guarantee the required data transmission success rate. Literature [13] discusses the scenario of using the SWIM system to collect biological information of whales. SWIM believes that the random mobility of sensors causes all sensor nodes to The probability of encountering the sink point is the same. In order to obtain the required data transmission success rate, the sensor node only needs to distribute copies of a certain number of data packets to the network. However, in actual applications, each sensor node and the sink node The probabilities of points meeting are not equal, resulting in low efficiency of SWIM.
In addition to the algorithms mentioned above, related research also includes literature [14,15]. The RED (replication-based efficient data delivery) strategy proposed in literature [14] consists of two parts: data transmission and message management. During data transmission, The calculation of transmission probability adopts an improved method based on historical records. Whenever message transmission occurs, the transmission probability of the node is increased. If no message communication occurs within a period of time, the value of transmission probability is appropriately reduced. Message management is based on The current transmission probability value of the node determines the optimal erasure coding parameters to improve the transmission success rate. However, the calculated values of the optimal erasure coding parameters in RED are not accurate [10], and a large number of small errors are transmitted in the network. Fragmented messages further aggravate the transmission energy consumption of the network. The FAD strategy [15] is formed on the basis of the RED strategy. FAD uses the same transmission probability calculation method as RED, and at the same time, based on the error tolerance of each message ( fault tolerance) value to manage the message queue. However, the transmission energy consumption of this mechanism is still large, and it ignores the survival time of the message, so it is possible that some messages whose transmission delay has exceeded the network delay tolerance limit continue to exist. The phenomenon of consuming network bandwidth and energy in the network. In addition, DTN technology is also widely used in self-organizing networks [16-19].
2 Network model and problem description 2.1 Network model This paper assumes that in the initial state, N sensor nodes are randomly distributed in an M×M two-dimensional square area A, and for simplicity, it is assumed that the only convergence point in the network is also deployed in the area A and fixed. The communication radius of all sensor nodes and sink points is R. In addition, it is assumed that the sensor network has the following properties:
The motion rules of all sensor nodes conform to the Random Waypoint motion model. To simplify the calculation, it is assumed that the motion speed of all nodes in the model is the same, which is V. The Random Waypoint motion model is described as: the sensor node randomly takes the starting point S and Destination point D, use constant speed V to move along a straight line from S to D, randomly select a time Tpause at D that belongs to (Tmin, Tmax) and remain stationary, thus completing a movement process. Use this destination point D as the starting point for the next movement. Starting point S, carry out the next movement process, and so on. All sensor nodes in the network follow the above movement process, and they are independent of each other. The movement process of the node is shown in Figure 2. • Since the convergence point is stationary, for all sensors For nodes, the location of the convergence point is known. • Each sensor node knows the destination point D of this node’s movement (before stopping). • Through the global positioning system GPS (global positioning system), each sensor node can know any time own current location. • Execute a time synchronization algorithm after all sensor nodes are deployed to keep all sensor nodes time synchronized. Of course, this algorithm does not have to be too precise. For example, some quasi-synchronous algorithms, due to the message timeout time in the DTN network It may be longer, such as minutes, hours or even days, so the synchronization interval can be larger, so the synchronization overhead can be assumed to be smaller.
2.2 Problem Description Compared with traditional sensor networks, DTMSN has the following characteristics: 1) Random mobility of nodes. Since sensor nodes or convergence points are placed on randomly moving objects, the topology of the network is dynamic. 2) Intermittent connectivity The connectivity of DTMSN is poor, and a certain sensor node is only occasionally connected to other sensor nodes. 3) Delay tolerance. Due to the intermittent connectivity between sensor nodes, the transmission delay of data in DTMSN is often high, so the application can tolerate Large data delay. In addition, the storage space of sensor nodes is limited, and this limitation has a significant impact on the performance of DTMSN. Because the data messages of sensor nodes must be stored in the queue for a long time before being transmitted to other sensor nodes or aggregation points. Time and how to design queue management strategies become bottlenecks.
In order to satisfy the above characteristics, an effective DTMSN routing algorithm needs to meet the following characteristics at the same time: 1) Dynamically select the sensor node that is most likely to be close to the convergence point at the next moment as the carrier of the message. If the next hop node is not selected correctly, then It will lead to excessive node energy consumption and a sharp decline in network performance. 2) Design appropriate message and queue management mechanisms. Reasonable queue management will definitely help improve the performance of the entire network. In the current research on DTMSN transmission algorithm, there have been The work cannot meet the above requirements at the same time. For example, the RED algorithm and the FAD algorithm both perform routing based on historical records, and the effectiveness of the history-based strategy depends on the motion status (speed, direction) of the node. Figure 3 is a system consisting of 8 sensors Schematic diagram of DTMSN consisting of nodes and a convergence point (the content in parentheses represents the transmission probability value calculated based on historical records, and the arrow represents the movement direction of the node), in which nodes located in the same circle can communicate with each other, and node 3 and node 3 4 has the highest transmission probability. According to the RED and FAD transmission strategies, it is considered that the communication frequency between these two nodes and the convergence point is the highest, so node 3 and node 4 are selected to become the next hop routing nodes. However, since the movement directions of the two nodes are both at this time. Departing from the convergence point, it is not effective for other nodes to pass data messages to these two nodes at the next moment. In order to solve the above problem, this paper proposes a dynamic data transmission strategy SRAD that meets the above two requirements at the same time. In the next section ,We will provide a detailed description of the SRAD ,strategy.