MD-MAC: A Distributed TDMA Protocol Based on Desynchronization for Multi-Hop Topologies

Round 1

Reviewer 1 Report

The authors have address my comments and is good for publication 

Author Response

Thank you very much for your previous suggestions. Best wishes

Reviewer 2 Report

This manuscript designs a distributed protocol, which does not rely on the prior knowledge of nodes, and neither central node to schedule global information. Through this method, they solve the problem of the communication delay between each node. Moreover, they propose a collision detection scheme to avoid collisions. However, the manuscript still exists following shortages: 1. Introduction line 26: Please describe the meaning of traditional network and FANET. 2. Please keep the context consistent: use “Section 2” to replace “Section ii”. 3. The nodes in the network are in a fully-connected network in line 91. What is the meaning of previous and after nodes in line 96? 4. The description of Figure 1 is unclear. 5. Please add simple annotation of Algorithm 1 for readers to understand. 6. In page 4: the computation complexity is proposed in line 140, but there is no detailed analysis of it. 7. Moreover, there are many spelling mistake, such as “relay” ->”rely” in line 76.

Author Response

Thank you for your comments. We will give out some proposals and explanations for solving these problems:

Response: 

1. What is the meaning of previous and after nodes in line 96?

For the DESYNC protocol, a period can be considered as a cycle. Therefore, according to the unique serial number assigned to each node, each node has neighbors on the ring, that is, the previous and the next node. For example, the pervious node of node b is node c, and the next node of node b is node a. The node also adjusts the phase according to the neighbors on the ring.

2. the computation complexity

For the computation complexity, no detailed calculations are made in this paper, but a general description is based on the transmitting data packets. The focus of this protocol is on continuously variable time slot length and adaptation to the dynamic topology. Thus, deleting the part of the computation complexity

 

Author action:

1. Introduce the traditional networks and FANET in the introduction section.

2. Modify the representation of the section number.

3. Add description of Figure 1 to the article.

4. Add  some comments to Algorithm 1.

5. Modify the spelling mistakes that appear in the paper.

Reviewer 3 Report

The paper proposes a distributed TDMA protocol based on desynchronization. It basically tries to adapt the well known DESYNC protocol to multi-hop ad hoc networks. I identified the following issues:

Manuscript's English is very poor. In my view the paper has to be rewritten from scratch. There are numerous examples of grammar mistakes, typos or badly-shaped sentences that make the manuscript almost unintelligible. Some examples from the first section (Introduction): a) in lines 17-19 there is a badly shaped sentence (“The importance …”); b) line 21: it is written “communicates” instead of “communicate”; c) line 25: after [5] we need a period instead of comma; d) line 25: it is written “important significance” which is confusing since “important” and “significant” are synonyms; e) lines 28-29: the sequence “which can not adopt traditional wireless communication protocol and need to be focused while designing the Mac algorithm.” is confusing; f) line 29: it is written “the protocol we learn” which seems strange; g) lines 44-45: the sentence “It can measure…” is confusing; and many-many more. Line 4: The sentence “Desynchronization is a biological term” is wrong. What the authors understand by “business demand” (line 27), “previous learning” (line 35), “midpoint of nodes” (line 96)? Lines 33-34: There is a need for a reference after the sentence “According to…”; Line 62: The sentence “Also, there must exist transmission collision while the topology changes.” Is confusing; Lines 81-86: The sections numbers must be Arabic numerals (for example instead of “ii” (line 81) must be written “2”); Lines 106: “the state of desynchronization is stable” is strange; In many places in the manuscript it is written “function” when referring to “equation”. Some examples are in line 128, line 154; line 156; etc. Lines 128-129: it is unclear how the order numbers are allocated in clockwise order; some details are needed. Some details about the scalability of the proposed method are needed; Lines 136-137: The sentence “When the transmission distance exceeds two hops, the information will be sent through nodes on routing.” Is confusing; Lines 142: a reference for the MTS algorithm is needed; Definition 1 is valid only for one-hop networks, but the paper is about multi-hop networks; The first sentence from section 3 is confusing; What the authors mean by “number of degree” (line 177)? Line 180: on what basis the lower limit for eta is set? Lines 182-183: This sentence is unclear. We cannot switch from one algorithm to another every time collisions are present. How can we do this? Also, there is a need for a reference concerning E-Mac; Lines 230-232: there are no details or an explanation why the authors say that the convergence to collision-free state is guaranteed? It is unclear how the desychronisation process is done when using multi-hop transmission. How the process is scheduled to obtain the time-slots for each node in this case? Lines 274: it is unclear how the uniform random punching number is obtained? What the authors mean by “range of neighbor nodes”(Table 1) or “topology size (line 331)? Lines 311-312: there is no proof for the sentence “When the length of slots varies less than 5% in the network, it means that the node gets the stable state”; The term “cycle of converge” is wrong (Fig.9);

Author Response

Thank you for your comments. We will give out some proposals and explanations for solving these problems:

Apologizing for grammar mistakes and badly-shaped sentences, and we have modified these mistakes which you mentioned in the paper.

Response: 

1.The understand by “business demand” (line 27), “midpoint of nodes” (line 96)

(1) business demand: in the multi-hop network, each node is self-governed and negotiates with nodes in the communication range. Thus, nodes in the dense network have more opportunities to transmit. Different transmission opportunity for nodes means business demand.

(2) midpoint of nodes: for the Desync protocol,  a period is considered as a ring, as shown in Figure 1 in the paper. Therefore, time slots occupied by nodes can be represented as phase. What the midpoint of nodes means that while the node meets the condition, it will adjust its phase to the phase midpoint of its neighbors.

2. transmission collision existing while the topology changes

The application scenario of the protocol is FANET, so topology changes are common. Since the networking mode we adopt is TDMA, the slot resources can be reused. However, while the topology changes, the nodes which were not in the communication range may be close to each other, which causes a transmission collision.

3. The sentence “When the transmission distance exceeds two hops, the information will be sent through nodes on routing.” is confusing

What I want to explain is that nodes in the multi-hop network can connect with others outside the communication range, so that in the protocol the situation we need to consider is that nodes negotiate with neighbors to occupy time slot.

4. Definition 1 is valid only for one-hop networks, but the paper is about multi-hop networks.

In Definition 1, the equation (2) means that the phase relationship between node and its neighbors, and equation (3) specifies the phase range of all nodes. For the multi-hop network, nodes can reuse the same time slot, so the sum of the phases is more than one cycle. Definition 1 for one-hop networks is presented as:

ΣΔΦ=1,  Σli≤T

5. the lower limit for eta/Lines 182-183: This sentence is unclear. We cannot switch from one algorithm to another every time collisions are present.

Although the slot length is dynamically changed to accommodate the topology in the protocol, the initial length is a fixed value. In order to calculate the fixed value, we propose two methods, which are the calculation using probability and the dichotomy. For the probability method, we determine the low limit for eta according to Equation (4) by limiting the probability of collision-free (Pi>95%).

These two methods do not switch to each other, but are determined to use by the topological density. When the topological density is small, namely Dmax is small, the low limit for eta can be obtained quickly by calculating the probability formula, and when the topological density is large, the probability formula error becomes large. So, we adopt the dichotomy to calculate.

6. how the desychronisation process is done when using multi-hop transmission.

The difference in multi-hop networks is that each node is only affected by neighbors within the communication range. As mentioned above, a period is considered as a circle and all nodes are assigned to the ring. But for a node, there are only neighbors in its view ring. Unlike a single-hop network, a node is affected by its one-hop neighbor and two-hop neighbor, so there is stale information. We analyze the different situations and add the timestamp mechanism to solve the error caused by the stale information, so that the nodes can adjust the phase (time slot length) one by one.

7. range of neighbor nodes / topology size

Range of neighbor nodes: in the simulation, we compare the convergence speeds of different protocols under different network sizes and network densities. Thus, the range of neighbor nodes means the range of maximum values of neighbor nodes of all nodes in the topology.

Topology size: The topology size represented in line 331 refers to the maximum value of the neighbor nodes of all nodes in the topology.

8. there is no proof for the sentence “When the length of slots varies less than 5% in the network, it means that the node gets the stable state”

This can be considered as a threshold for the node to reach a steady state, so each node will calculate the adjustment length according to the previous state each period. We specify that while the phase adjustment value is less than 5%, the system is considered stable.

Author action:

1. Modifying the grammar and sentence errors which appear in the article based on the reviewer's comments.

2. Supplying the unclear part mentioned by the reviewer.

3. Modifying the error expression of Figure 9 and 10.

Specific modifications have been highlighted in the paper.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The authors have significantly improved their manuscript. They have successfully solved all my comments and concerns.

Author Response

Thank you very much for your comments.

Best wishes to you

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This paper introduces a distributed TDMA MAC protocol for WSNs. The arguments carried in the paper are convincing, but I see some very critical issues regarding the protocol design and evaluation.

 

1. The scheme does not well-consider clock drift, which is common on low-cost WSN nodes. Many well known schemes already consider this in the time sync phase and should be considered when designing a protocol for multihop WSNs

2. Evaluation is very weak. First, the channel/noise environment, in which the protocol was tested is not specified. Second, only a single topology is used, despite the need for generality in MAC protocols. Third, I suggest the performance comparisons to be against many of the MAC protocols introduced in top conferences such as SenSys and IPSN. 

Author Response

Response to reviewer 1

Thank you very much for your helpful suggestions.

Q1. The scheme does not well-consider clock drift, which is common on low-cost WSN nodes. Many well known schemes already consider this in the time sync phase and should be considered when designing a protocol for multihop WSNs

Response(1): Clock drift problem:

The clock drift is an important part for WSNs. In the paper, we adopt the MTS(Maximum Time Synchronization) protocol to synchronize time, because of the characteristic of  distribution, a-synchronization and robustness. The clock messages are sent to neighbors when the node reaches the firing point. Nodes can adjust their logical clocks and the logical clock of all nodes will equal to the maximum hardware clock eventually. It is useful for the low-cost WSN nodes.

Q2. Evaluation is very weak. First, the channel/noise environment, in which the protocol was tested is not specified. Second, only a single topology is used, despite the need for generality in MAC protocols. Third, I suggest the performance comparisons to be against many of the MAC protocols introduced in top conferences such as SenSys and IPSN.

Response(2): The weakness of evaluation part:

The compared protocol we choose is the Z-Mac which has been proposed in Sensys 05’. The main idea of this protocol is to combine the TDMA and CSMA. When the node is in the high contention level, it occupies slot in the TDMA way, and in the low contention level, it will compete with its neighbors. Thus, we compare our protocol with Z-Mac in the throughput and convergence speed. The topology used for simulation is randomly generated in each iteration because in the program the node will be distributed randomly in each iteration. And the topology shown in figure 7 is an example in our simulation.

Author action:

In section five, we add the table of simulation parameters including link capacity, SINR, average transfer delay and so on. We have added the performance simulation of Z-Mac comparing with our protocol, and the simulation results have been given in the revised manuscript. The description of time synchronization protocol has been added in section 2.2. And the further study about time synchronization will be the point for the later work.

Reviewer 2 Report

This is a good paper. However I would like the authors to explain better the transmission range or coverage of each node, specially for the cases in which they mention the 2-hops situation. The hidden and expose station problems are not mentioned and neither is said about the optimal frame length in the TDMA access. Basically nodes distant away as to not listen to each other may anyway interfere over another one with capacity to listen to both of them, and if they are far enough as to not have this problem they may use the same slot to transmit. 

Another aspect not mentioned in the paper is the energy demand during the synchronization phase and how nodes may end-up routing messages to neighbor nodes if they are not directed to them. I mean, node A wants to send a message to node D but for this nodes B and C has to act as intermediate steps. This compromise both memory and energy in this action and besides compete with the own generated messages. It would be nice to have a paragraph on this aspect.

Along the paper there are minor typos  that with a good proofreading would probably be solved.

Author Response

Response to reviewer 2

Thank you for your comments. We will give out some proposals and explanation for solving these problems:

Response(1) The hidden and expose station problems:

In the protocol we proposed, all nodes will jump until making sure the phase of their previous and next neighbors, and keep the information fresh by time stamp rule. This process can be seen as an allocation Mac protocol, not a competition one. Thus, we can neglect the hidden and expose station problems.

Response(2) the frame length:

The period length is constant in the simulation, and we set it 100. The initial frame length is calculated by the function(4) given in the paper or the dichotomy model. And the optimal length in the stable status is determined by each node and its neighbors, because the firing point of each node will keep in the midpoint of the previous and next neighbors. A more detailed analysis will be the point for the later work.

Response(3) the transmission range:

In this paper, we control the number of neighbors in the simulation through changing the communication distance which is set from 60 to 120m. Thus, the transmission range is different in the different scenarios.

Response(4) energy, memory and the routing problem:

When a node transmits to the node two-hop distance away, the node on routing will forward the packet through the target node address. The paper considers the level-two (Mac Layer) problem. Thus, the forwarded message in the level-three (Network Layer) is the same as the one transmitted by a node in the level-two. The address of the target node is stored in the frame head, so each node on routing will recognize the address and send it to the target node. In addition, each node stores information including address, clock, data and so on in the range of two hops, and the computation complexity is low. Thus, the memory and energy each node needs are in a low level.

Author action:

In section 2.2, we have provided the explanation about energy, memory and routing problems. We add the simulation parameters including the transmission power in section 4.

Round 2

Reviewer 1 Report

Thank you authors for addressing my comments well. However, I still question the contents of the evaluation section and the results. First of all, there is no justification of the simulation parameters in Table 1. Furthermore, an important aspect of WSN protocols, which we agree with MD-MAC is designed for, energy/radio duty cycling is an important issue. I do not see any discussion/evaluation on this issue. 

 

Minor issue, but the captions for Figures 9 and 10 are not at all informative.