KR102180320B1 - Method for efficient convergecast scheduling in multihop network - Google Patents

Abstract

Disclosed is an efficient convergedcast scheduling method in a multihop network. The method includes the steps of determining, by a path computation element (PCE), a set of parent nodes based on information of a child node, and giving priority to each of the parent nodes by the child node, The child node transmits a data frame to all the parent nodes in an anycast method, and a parent node having the highest priority among the parent nodes and correctly receiving the data frame transmits a response. Includes steps.

Description

Efficient Convergencecast Scheduling Method in Multi-Hop Network {METHOD FOR EFFICIENT CONVERGECAST SCHEDULING IN MULTIHOP NETWORK}

An embodiment according to the concept of the present invention relates to an efficient convergedcast scheduling method in a multi-hop network, and in particular, to an efficient scheduling method for using opportunistic characteristics of the MAC layer.

Recent advances in technology have enabled the creation of smart objects that can be interconnected to create a new Internet of things (IoT). The IEEE 802.15.4 working group proposed an amendment in 2012 to enable energy efficient networking in the industrial market. Since transmission is often predictable, the timesloted channel hopping (TSCH) mode has proposed allocating time slots combined with a slow channel hopping strategy to improve reliability. For energy efficiency, TSCH implements implicit synchronization. Any data or control packet can be used at the receiver or transmitter to calculate the clock drift. This strategy is particularly efficient for periodic and/or predictive transmission.

In a multihop network, a centralized path computation engine (PCE) can calculate the path to be used for each flow. Alternatively, routing protocols such as RPL (IPv6 Routing Protocol for Low Power Lossy Networks) can configure paths distributedly.

For traffic in the form of convergecast, most approaches construct a directional aperiodic graph based on border routers (ie, gateways of the Internet). Then, each node has a set of parents that make up the next hops towards the border router. However, currently RPL only selects one parent node to carry all traffic, and other nodes are used for backup purposes.

Registered Patent Publication No. 10-1363219 relates to a communication method using a multi-hop wireless network using an opportunistic routing technique. A destination node opportunistically receives a data packet, and the reception result is transmitted on a transmission path of the data packet. Disclosed is an invention for improving the throughput of a network by notifying other nodes to prevent unnecessary transmission of already-transmitted data packets.

However, in the prior art document, since one destination node opportunistically receives data packets transmitted by each of the plurality of nodes from the source node and nodes located on the path from the source node to the destination node , There is a problem that reliability is poor.

Registered Patent Publication No. 10-1363219

The present invention has been devised to solve the above problems, and an efficient convergencecast scheduling method in a multi-hop network of the present invention aims to effectively select parent nodes and transmit traffic in consideration of channel quality.

In an efficient convergedcast scheduling method in a multi-hop network of the present invention for achieving the above object, a path computation element (PCE) determines a set of parent nodes based on information of a child node, and , The child node giving priority to each of the parent nodes, the child node transmitting a data frame to all the parent nodes in an anycast method, and the most of the parent nodes And transmitting a response by a parent node having a high priority and correctly receiving the data frame.

In addition, the step of transmitting the response by the parent node includes: performing a continuous clear channel assessment (CCA) by the parent node at turn-around time intervals, and the parent node is different from And transmitting, by the parent node, if a response is not received from the parent nodes.

In addition, the step of transmitting the data frame and the step of transmitting the response are performed every timeslot of a predetermined length.

In addition, the timeslot includes a data frame period for transmitting the data frame and a response period for transmitting the response.

In addition, the determining of the set of parent nodes is determined based on an energy budget of the child node and a channel state of the child node and the parent node.

In addition, the channel state is a first state when the signal-to-noise (SNR) measured in the radio link between the child node and the parent node is lower than a threshold value, otherwise The state of the channel has a second state.

In addition, the energy budget means transmission energy of the child node in a specific channel state when transmitting the data frame to the parent node, and satisfies Equation 1 below.

[Equation 1]

는 상기 자식 노드가 상기 부모 노드로 상기 타임슬롯 t 동안 상기 데이터 프레임을 전송할 수 있음을 의미하고,

는 상기 자식 노드와 상기 부모 노드 사이의 무선 링크의 채널 상태이고, N₀는 부가적인 가우시안 잡음(Gaussian noise)이고, r은 전송률이고, W는 대역폭이다.here,

Means that the child node can transmit the data frame to the parent node during the timeslot t,

Is a channel state of the radio link between the child node and the parent node, N ₀ is an additional Gaussian noise, r is a transmission rate, and W is a bandwidth.

As described above, the efficient convergedcast scheduling method in a multi-hop network according to the present invention uses an opportunistic multi-parent node access, thereby reducing energy consumption of the network and increasing reliability.

A detailed description of each drawing is provided in order to more fully understand the drawings cited in the detailed description of the present invention.
1 shows a super frame structure according to an embodiment of the present invention.
2 shows the DODAG structure of RPL according to an embodiment of the present invention.
3 shows a time slot structure for radio links in the structure shown in FIG. 2.
4 is a conceptual diagram illustrating a channel state according to an embodiment of the present invention.
5 is a flowchart illustrating a scheduling method according to an embodiment of the present invention.

Low power and lossy networks (LLNs) aim to integrate smart objects into the Internet of thins (IoT). IEEE 802.15.4-TSCH is a promising standard for the current link layer. While scheduling transmission and implementing slow channel hopping to improve stability, the routing layer concentrates on constructing distributed paths for a set of small destinations (ie, convergecast) In this specification, the opportunistic characteristics of the MAC layer are discussed. We propose an efficient scheduling policy to use: A single transmission is received by opportunistically determining which one of the set of next hops a packet will be transmitted to. The MAC and routing layers take into account this opportunistic forwarding, and thus reliability and energy efficiency. Consider the optimal scheduling policy for each packet The simulation results prove the validity of the proposed policy The energy consumption per packet is reduced by effectively selecting the parent nodes (ie, next hops) and carefully considering the channel quality. In addition, it improves reliability The network can also use unreliable radio links, where only one of the next hops receives and forwards a packet This scheduling policy is generally the IEEE 802.15.4 e-TSCH path. It can be implemented in the path computation engine (PCE).

Here, in order to modify the MAC forwarding strategy to enable opportunistic forwarding, it is proposed to improve reliability by using all parent nodes. Currently, 6TiSCH focuses on unicast transmission using the concept of a track. Here, it aims to reduce energy consumption and increase reliability by utilizing several parent nodes in parallel with the opportunistic version of IEEE 802.15.4-TSCH. Typically, such a solution can be implemented as a 6TiSCH with a centralized controller (PCE). In order to propose an efficient schedule, it is necessary to determine the optimal transmission power. In fact, using maximum power can improve reliability, but it can also increase energy consumption. In addition, the channel quality changes over time due to fading. Thus, significant energy conservation can be achieved by scheduling transmissions during times when the channel is of good quality. Here, it is possible to improve end-to-end reliability and global energy consumption in consideration of a scheduling problem in a multi-hop convergecast network. In particular, multiple parent node opportunistic scheduling is used in the MAC layer to cope with individually defective links.

In IEEE 802.15.4-TSCH, a schedule is set so that at the beginning of each timeslot, each node knows whether it must be awake to receive or transmit a frame. A cell represents a transmission opportunity represented by a timeslot and a channel offset.

1 shows a super frame structure according to an embodiment of the present invention. In general, IEEE 802.15.4 supports a star structure, in which communication is performed through a coordinator as a medium. The coordinator periodically transmits a beacon, and a node belonging to one cell may transmit data during a superframe, which is a transmission period between the beacon and the beacon.

Referring to FIG. 1, a slot frame includes a specific number of timeslots, and is repeated over time. A slot is assigned to one transmitter or shared between a group of interfering nodes. The radio link (C→PANc) reserves two cells (timeslots 5 and 7) for transmission.

TSCH uses centralized scheduling using a path computation engine (PCE). PCE refers to a central entity that calculates paths used for each flow and cells used for each radio link. In this case, the 6TiSCH structure is an extension of a deterministic networking architecture adopting a software-defined networking approach.

Schedules can also be distributed and determined hop by hop to negotiate a group of cells using the 6top protocol. Since each node pair automatically selects a cell to be used, interference may occur. Some mechanisms are needed to monitor and adjust the local schedule.

Interference can occur in the following cases.

Internal interference occurs when the same cell is assigned to a pair of interfering transmitters. Because media access is not implemented, collisions occur in all cells. Here, we assume that the interference topology is a priori, and our scheduling algorithm prevents allocating the same cell to two different interfering transmitters. So we ignore this internal interference.

External interference can occur in other wireless technologies (Bluetooth, Wi-Fi, etc.) because ISM band is used [12, 13]. However, channel hopping is particularly effective in reducing the number of retransmissions by mitigating this kind of external interference [14]. Consider that external interference negatively affects the packet error rate (PER). When there is external interference, more cells should be allocated. Moreover, it focuses on long-term performance. That is, the external interference is sufficiently stable to be estimated through PER and is considered in the scheduling process. We investigate the effect of external interference in the simulation.

2 shows the DODAG structure of RPL. Since IEEE 802.15.4e-TSCH-based devices use a time-division multiplexing (TDM) method, time synchronization between nodes is important. Therefore, time synchronization is achieved based on the RPL route, and a path setting operation based on RPL DODAG (destication oriented directed acyclic graphs) is performed.

Referring to FIG. 2, it is notified that a parent node having a higher priority has received a packet first. They use the MAC layer where they have the opportunity, where a single transmission is sufficient to deliver the packet to all parent nodes. Opportunistically, with the highest priority, the parent node receiving the packet announces reception and forwards the packet.

In this specification, a method of discovering how opportunistic forwarding can improve reliability while limiting energy consumption in the MAC layer is proposed. These multi-parent node functions can be incorporated into these scheduling algorithms. TSCH is modified to perform opportunistic forwarding. That is, the child node transfers the data frames to any parent node among the parent nodes by anycast.

In addition, it is an object of the present invention to implement the scheduling method of the present invention in a path computation element (PCE). This makes it possible to completely reuse the 6TiSCH protocols.

3 shows a time slot structure for radio links in the structure shown in FIG. 2. Referring to FIG. 3, it is assumed that in radio links (A,B) and (A,C), a parent node B is preferred to a parent node C. In the first timeslot, parent node B will respond first (ack(B)), and C will stop listening to the medium. In the second timeslot, B did not correctly receive the packet, and C will respond after a sufficient delay (ack(C)). Transmission fails only if all parent nodes are unable to decode the packet.

Specifically, for opportunistic TSCH, some parent nodes must wake-up at the start of the assigned cell. The child node transmits the data frame in anycast method, and the parent node that correctly receives the packet with the highest priority must send a response.

After the data packet part of the timeslot, the remaining part of the timeslot is allocated for opportunistic responses. Child nodes are given priority among parent nodes. If the parent node does not hear a response from the other parent nodes for a sufficiently long time, the parent node decides to forward the packet.

More specifically, the k-th parent node starts k consecutive clear channel assessments (CCA) at turn-around time intervals. Here, the turnaround time means the maximum time for switching from the reception mode to the transmission mode. Parent nodes are strictly ordered, so the parent node stops deferring as soon as it receives a response.

To reduce the overhead, we limited the number of possible active parent nodes in a given timeslot to the upper limit that can be allowed. In fact, four parent nodes are sufficient to provide a certain variety. For a given CCA period (128 μs) and turnaround time (210 μs), approximately 2 ms is allocated for response at the end of the timeslot.

Parent nodes must listen to each other to hear responses from other parent nodes. To construct a schedule, a node must report the link quality to its neighbors. Therefore, the path computation engine (PCE) will select a set of parent nodes with a very small PER in a given timeslot. In addition, if the CCA fails, i.e., there is no need to decode the response packet, and only needs to detect radio activity above the CCA threshold, the parent node will consider the packet to be answered.

A convergedcast traffic problem is considered in wireless sensor networks including one border router and N sensor nodes. Sensor nodes regularly report measurement values (eg pollution, noise, gas or water consumption) to the root node. All sensor nodes are powered by batteries with limited energy. In an embodiment of the present invention, since the DODAG structure is configured using a routing protocol such as RPL, each node can transmit data to a route using their parent nodes. An object of the present invention is to minimize the transmission energy of the entire network in an environment in which the total amount of received data is maximized by determining parent nodes associated with a series of active nodes allowed to transmit data. For convenience of explanation, symbols in this specification are summarized in Table 1.

System model

i ∈ N represents the index of the sensor node, and Pi represents the set of parent nodes. When the two nodes are within a radio range of each other (ie, when power consumption for decoding a frame is sufficient), the two nodes can communicate with each other. We consider the structure of a slotted time frame used in IEEE 802.15.4-TSCH. Time is divided into equal cycles (slot frames). As described in FIG. 1, each cycle consists of T timeslots, and the timeslot is a basic unit time for transmission. At the beginning of each cycle, the scheduling policy determines the set of nodes allowed to transmit, based on the energy budget and channel conditions of each sensor node in the cycle.

Channel model

Channel conditions can be modeled using a finite-state markov channel (FSMC) model. Each node may use some allocated beacon frames to evaluate the signal-to-noise (SNR) on the radio link. Here, the range of possible SNR values is divided into equal sections, and each section represents the state of the Markov chain.

라 한다.

는 노드 i의 채널 상태를 나타내는 벡터이고, 여기서

는 노드 i와 부모 노드 j 사이의 무선 링크의 채널 상태들이다. 이러한 상태들 S_i(t) 중에서 센서 노드들의 채널 상태들은 Markov 프로세스를 따른다. t+1에서 링크

의 연속적인 상태들

은 전이 확률

에 의해 결정된다. The set of states

It is called.

Is a vector representing the channel state of node i, where

Are channel conditions of the radio link between node i and parent node j. Among these states _Si (t), the channel states of sensor nodes follow the Markov process. link at t+1

Successive states of

Is the probability of transition

Is determined by

This transition probability is estimated through the probability density function (PDF) of the SNR. Each Markov state represents an interval of SNR. From this function we can estimate the state transition probability in the Markov model. This probability is equal to the probability that there is an SNR within a given interval.

Interference model

는 노드 i의 활성화 상태를 나타내는 벡터이다. 여기서 만약 노드 i가 부모 노드들

로 데이터를 전송하면

이다. 본 명세서에서, 기회주의적 멀티-부모 노드 접근을 사용할 경우, 각 노드는 도 3에 도시된 타임슬롯 구조를 이용하여 동시에 부모 노드들로 데이터를 전달할 수 있다.

Is a vector representing the activation state of node i. Where if node i is the parent nodes

When you send data to

to be. In the present specification, when the opportunistic multi-parent node access is used, each node may simultaneously transmit data to the parent nodes by using the timeslot structure shown in FIG. 3.

To determine which radio links are scheduled at the same time, we use a contention graph model. Each vertex of the contention graph represents a radio link, and when a pair of radio links interfere with each other, the pair of radio links are neighbors (ie, edges exist) in the contention graph. We use the interference range model here to construct the competition graph. If the distance between the two nodes is shorter than the interference range, the two nodes cannot be scheduled at the same time, and an edge exists in the contention graph. Other interference models are also relevant, but should only be mapped to the contention graph.

를 부모 노드들 Pi 에 관한 노드 i 의 실현 가능한 활성화 벡터라고 하고,

는

의 모든 노드들이 동시에 활성화될 수 있는 실현 가능한 집합이라고 하자. G_C 는 경쟁 그래프를 나타낸다. 실현 가능한 집합은 독립적인 집합, 즉, 임의의 두개의 꼭지점들 사이의 엣지가 경쟁 그래프에 존재하지 않는 무선 링크들의 집합에 해당한다는 것이 명백하다. 독립적인 집합을 찾는 것은 NP-hard 문제로 널리 알려져 있다. 그러나, 복잡성을 줄이기 위한 몇가지 기술이 있다. 게다가, 특정 네트워크 토폴로지를 위해, 우리는 초기에 한번만 독립적인 집합을 계산해야 한다.

Is the feasible activation vector of node i with respect to the parent nodes Pi,

Is

Let's say it is a feasible set in which all the nodes of can be activated simultaneously. G _C Represents the competition graph. It is clear that the feasible set corresponds to an independent set, i.e., a set of radio links whose edges between any two vertices do not exist in the contention graph. Finding an independent set is widely known as the NP-hard problem. However, there are several techniques to reduce the complexity. Moreover, for a particular network topology, we have to compute the independent set only once initially.

에 의해 표현된다. 특히 이 지표 함수는 다음과 같이 정의된다:In this specification, a heuristic algorithm is adopted to determine a subset of independent sets. It has been found that under this heuristic algorithm the rate is very close to that calculated using all possible transfers. The success/failure of packet delivery from node i to parent node j is an indicator transmission random variable depending on channel conditions and schedule policy

Is represented by Specifically, this indicator function is defined as follows:

는 확률 변수이고, 주어진 채널 품질 하에서

가 주어진 조건부 기대(conditional expectation)는 성공 확률과 동일하다. 그러므로, 특정 재널 상태들

에서 활성화 벡터,

는 결정 함수 0/1이 될 수 있다.

Is a random variable, and under a given channel quality

The conditional expectation given is equal to the probability of success. Therefore, certain channel states

Activation vector in,

Can be a decision function 0/1.

이다. 상태 i에서 상태 j로의 전이 확률은 p^i,j 에 의해 결정된다. 그렇지 않으면, 각 상태는 성공 전송 확률들

과 연관된다. 무선 링크(i,j) 상의 지표 전송 I_i,j 는 확률 p_i를 갖는 랜덤 프로세스에 따라 슬롯을 통해 진화한다. 이 예에서, 채널 상태가 B이면 전송은 실패하고(p_B=0) 채널 상태가 G이면 성공적인 전송 가능성은 p_G가 된다. Embodiment: FIG. 4 is a conceptual diagram for describing a channel state. In particular, FIG. 4 shows a Markov channel in two states called a Gilbert-Elliot channel. Referring to Fig. 4, each state in this channel model corresponds to a specific channel quality that is very good or completely bad. If the measured SNR is lower than the threshold value, the channel is displayed as "BAD" (B), otherwise the channel is displayed as "GOOD" (G). Channel status is

to be. The probability of transition from state i to state j is p ^i,j Is determined by Otherwise, each state is the success transmission probabilities

Is associated with The indicator transmission I _i,j on the radio link (i,j) evolves through the slot according to a random process with probability p _i . In this example, if the channel state is B, the transmission fails (p _B = 0), and if the channel state is G, the probability of a successful transmission is p _G.

Energy model

The common assumption is that communication radios, which contain energy for transmitting packets and energy for receiving packets, are a source of energy consumption. However, the overall energy consumption for receiving a packet is fixed and does not depend on channel conditions and schedule policy. Therefore, it focuses on optimizing the transmission energy.

는 패킷을 부모 노드 j에게 전송할 때, 특정 채널 상태

에서 노드 i의 전송 에너지라 하자. 이 에너지는 잘 알려진 Shannon 용량 공식을 이용하여 계산될 수 있다:

Is the specific channel state when transmitting the packet to the parent node j

Let be the transmission energy of node i at. This energy can be calculated using the well-known Shannon capacity formula:

[Equation 1]

는 채널 조건 s_i,j에서 노드들 i와 j 사이의 채널 이득이다. 만약 x_i,j(t) = 0 이면,

이므로, 할당 전력은 활성화 벡터의 함수이기도 하다.

를 시간 t에서 노드 i의 전송 에너지라 하자. 모든 패킷들이 모든 부모 노드들로 도달할 수 있는 충분한 전력으로 전송되도록 보장한다. 표기의 편의를 위해, 나머지 논문에서, 우리는

,

대신에

, 및

를 사용한다.

에 대한 의존성은 함축적이다.Here, N ₀ is the additional Gaussian noise, r is the transmission rate, and W is the bandwidth.

Is the channel gain between nodes i and j under the channel condition s _i,j . If x _i,j (t) = 0,

Therefore, the allocated power is also a function of the activation vector.

Let be the transmission energy of node i at time t. It ensures that all packets are transmitted with sufficient power to reach all parent nodes. For ease of notation, in the rest of the paper, we

,

Instead of

, And

Use.

The dependence on is implicit.

Formulate the problem

Each node i is associated with a utility:

[Equation 2]

는 각 노드에 대한 처리량 최대화 및 에너지 소비 최소화라는 두 오브젝트들 사이의 트레이드 오프를 제어한다.It describes the balance between throughput (ie, number of packets received) and allocated energy. parameter

Controls the trade-off between the two objects, maximizing throughput and minimizing energy consumption for each node.

의 값이 시뮬레이션을 통해 정해지므로 불명확하다. 따라서 하기의 [수학식 3]과 같이, 각 노드에 대한 처리량 최대화 및 에너지 소비 최소화의 비율(ratio)로 정의된 새로운 유틸리티 함수를 생성할 수 있다.However, the utility function of [Equation 2] is a parameter

It is unclear because the value of is determined through simulation. Accordingly, as shown in Equation 3 below, a new utility function defined as a ratio of maximizing throughput and minimizing energy consumption for each node can be generated.

[Equation 3]

를 제거함으로써, 스케줄링의 전처리과정에서 시뮬레이션을 통해 정해지는 파라미터

에 의해 지역 최적값(local optima)을 놓칠 수 있는 가능성을 없앨 수 있다.The new utility function in Equation 3 is an unclear parameter

By removing the parameters determined through simulation in the preprocessing of scheduling

The possibility of missing the local optima can be eliminated.

Consider the long-term network performance for one cycle length consisting of T timeslots. Network performance is characterized by the networkwide utility of all sensor nodes over T time slots. Each sensor node i has D _i packets to be transmitted during one cycle. You need to maximize your total utility:

[Equation 4]

Depending on the constraints on energy consumption, the transmitted packets are:

[Equation 5]

[Equation 6]

.

.

Equation (5) limits the energy budget of each sensor node within one cycle. Equation (6) ensures that each sensor node i transmits D _i packets during one cycle. This problem falls into the category of nonlinear integer programming known as NP-hard. Otherwise, it is difficult to find an optimal solution because the channel state of each user is uncertain and the complexity of the problem spreading over T time slots and N nodes gradually increases.

5 is a flowchart illustrating a scheduling method according to an embodiment of the present invention. Referring to FIG. 5, at the beginning of each cycle composed of T timeslots, a path computation element (PCE) determines a set of parent nodes corresponding to child nodes (S100).

The child node gives priority to each of the parent nodes (S200).

The child node transmits a data frame to all the parent nodes in an anycast method (S300).

A parent node having the highest priority among the parent nodes and correctly receiving the data frame transmits a response (S400).

Although the present invention has been described with reference to the exemplary embodiment illustrated in the drawings, this is only exemplary, and those of ordinary skill in the art will appreciate that various modifications and other equivalent embodiments are possible therefrom. Therefore, the true technical protection scope of the present invention should be determined by the technical idea of the attached registration claims.

Claims (7)

At the beginning of each cycle consisting of T timeslots, determining, by a path computation element (PCE), a set of parent nodes corresponding to the child nodes;
Giving priority to each of the parent nodes by the child node;
Transmitting, by the child node, a data frame to all the parent nodes in an anycast manner; And
And transmitting a response from a parent node having the highest priority among the parent nodes and correctly receiving the data frame. The method of claim 1,
The step of transmitting the response by the parent node,
Performing, by the parent node, a continuous clear channel assessment (CCA) at turn-around time intervals; And
And transmitting, by the parent node, a response when the parent node does not hear a response from other parent nodes. 2. An efficient convergedcast scheduling method in a multi-hop network comprising: The method of claim 1,
The transmitting of the data frame and the transmitting of the response are performed every timeslot. The method for efficient convergedcast scheduling in a multi-hop network. The method of claim 3,
And the timeslot includes a data frame period for transmitting the data frame and a response period for transmitting the response. The method of claim 1,
The step of determining the set of parent nodes,
And determining based on the energy budget of the child node and channel states of the child node and the parent node. 15. An efficient convergedcast scheduling method in a multi-hop network. The method of claim 5,
The channel state is a first state when a signal-to-noise (SNR) measured in a radio link between the child node and the parent node is lower than a threshold value, otherwise the channel state The state of is an efficient convergedcast scheduling method in a multi-hop network, characterized in that it has a second state. The method of claim 5,
The energy budget means transmission energy of the child node in a specific channel state when transmitting the data frame to the parent node, and efficient convergedcast scheduling in a multi-hop network, characterized in that it satisfies Equation 1 below. Way.
[Equation 1]

here,

Means that the child node transmits the data frame to the parent node during the timeslot t,

Is the channel state of the radio link between the child node and the parent node, N0 is additional Gaussian noise, r is a transmission rate, and W is a bandwidth.

Images