======= Review 1 ======= *** Familiarity: Rate your familiarity with the topic of the paper. Familiar (I am well aware of research work in this topic) (3) *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). accept: top 20% of all submissions, but not top 10% (4) *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. The main novelty of this paper is a distance-based carrier sensing mechanism to allow VANET nodes transmit more aggressively adaptive to the relative distance from/to surrounding nodes. This CS mechanism is combined with Reverse back-off scheme and a simple transmission power control; and yielded good performance improvements shown by simulations. The design concept is backed by an analysis on different types of collisions. This is a well-written paper with thorough background and related-work studies, important problem formulation and solid simulations. Building and maintaining a signal propagation model based on previously received frames' location info + a new frame PLCP's tx_power field is an interesting idea. However, in order to have an accurate distance estimation from the propagation model, I believe the model parameters should be kept per each transmitter, which requires the PLCP header to have transmitter ID field in addition to the power field. Despite the weakness, this paper can bring up interesting discussions among conference audiences. Thus, my rating is between accept and weak-accept, slightly leaning toward accept. *** Strengths: What are the major reasons to accept the paper? [Be brief.] + good problem motivation + thorough background and related work studies + reasonable analytic studies and strong simulations *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] The used propagation model is too simple and can't work well. The authors are suggested at least to try to fit real link measurement data to multiple regression models. The distance estimation model can't work well as it is because each transmitter-receiver pair link can exhibit very different characteristics. Even though we have the receiver in common, different transmitters and the prop paths from them to the same receiver can result in very different prop model parameters as studied in cellular and mesh networks. The use of directional antennas or free-space optical communications for VANET will make this per-transmitter model more needed. *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. Linear topology model doesn't accurately reflect multi-lane highways with many intersections. Though the model is not directly used for protocol design in the paper, there should be a good model to set right parameters and thresholds, which are not seriously considered in the paper. Several notations and terms are used without right description in right locations. ex) P_i and X. The argument of using ref [11] to support the RSSI-based power/distance estimation in varying channel is weak because ref. [1] is about radar application where a lot of continuous measurements of reflected OFDM signal is possible. 802.11 a/b/g/n/ac standards and chipsets clearly separate PHY and MAC by allowing the frame is sent to MAC after the last bit of PLCP + CRC is received at PHY. This separation easies the hardware and software design. I haven't checked 802.11p standard but I'm quite sure 11p follows the same. Allowing 802.11p PHY to access MAC or facilities layer would be an interesting topic to consider. Another option would be adding location or transmitter ID field into PLCP header but this will require solid evaluation to show the performance gain against the additional overhead in PLCP. ======= Review 2 ======= *** Familiarity: Rate your familiarity with the topic of the paper. Expert (I conduct(ed) active research work in this topic) (4) *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). accept: top 20% of all submissions, but not top 10% (4) *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. The authors propose a modified CSMA for the broadcasting of safety beacons in vehicular networks. Their solution operates by letting a vehicle evaluate locally if the simultaneous transmissions of its beacon and that of another vehicles may coexist given their distance. The paper has good technical depth and covers the major issues (such a power control and RSSI-based range estimation. Part of the proposed approach (the one on enhanced backoff mechanisms was already published by the same authors at ICC). *** Strengths: What are the major reasons to accept the paper? [Be brief.] The solution has good technical depth and the authors carry a good combination of analysis and simulation to make their case. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] Some evaluation comes somewhat short of what could have been expected (see below) *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. - The authors acknowledge that accurate RSSI-based range estimation requires extensive training and profiling, and cite previous literature work. However, some (minimal) performance evaluation on the impact of distance under- (or over-)estimation in SR-CSMA could have been useful. - A simulation-based validation of numerical results in figs. 2-4 should have been provided. ======= Review 3 ======= *** Familiarity: Rate your familiarity with the topic of the paper. Some knowledge (I am marginally aware of research work in this topic) (2) *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). weak accept: top 30% of all submissions, but not top 20% (3) *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. This paper proposes a modified CSMA MAC protocol for VANETs. The goal is to come up with a MAC protocol that can handle the significant network congestion levels that can arise in dense (congested) traffic conditions, due to the broadcast nature of safety messages and the large range of 802.11p. The authors perform a nice review of related (research and standardization) work and argue that, while power control has been considered correctly to reduce the range when car density increases, the effect of the carrier sense threshold has not been equally considered. Additionally, the authors observe that, since safety messages are more crucial for closeby neighbors than for the ones at the edge of the transmission range, it might make sense to define a "crucial range" around each node (quite shorter than the transmission range) and adjust the carrier sense threshold to ensure collisions occur only with far away nodes (not ones within each others safety range). In principle, this attempts to "pack" as many parallel transmissions as possible when congestion becomes very high. The authors then propose a MAC protocol to achieve this, that they supplement with some additional (existing) ideas for power control, and backoff. They evaluate the proposed scheme using real city maps and realistic traffic models. While there is nothing really technically challenging in this paper, the paper gives a good overview of related work, and makes a reasonable case that the proposed modification is new and make some difference. The simulation-based evaluation is also solid. On the other hand, I have two main concerns that weaken my impression of the paper: 1) It seems that the safety distance should be quite smaller than the transmission distance, e.g. 100m. This seems too small. It is essentially a line-of-sight distance for most scenarios, making the added value of VANET safety messages (meant to extend the horizon of the driver) less interesting. 2) The demonstrated performance benefits end up being less than 10% (at a 95% confidence interval). This does seem like a considerable improvement, making one less motivated to apply the proposed modifications and live with the somewhat increased complexity. *** Strengths: What are the major reasons to accept the paper? [Be brief.] The paper does a good job to position itself among related work, and the contribution, while incremental, is well explained. The proposed protocol seems to cover all cases and a realistic evaluation shows some performance gains. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] There are some practical concerns as to whether the safety distance applied in the end is not too small to be interesting for a VANET application. The contribution is incremental, and the performance improvement small. *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. All analysis assumes a "binary" approach to interference. Specifically, X decides to transmit if it doesn't interfere with Z. Then why decides to transmit if it doesn't interfere with Z either. But if both X and Y end up transmitting in parallel, they might create enough interference to kill Z's reception. The influence of aggregate interference is ignored throughout the analysis, and in the protocol design, since it requires the knowledge of the received power (or SINR) at the receiver, which is not part of the scheme. While the authors make a quick comment on this, saying that topology of VANETs doesn't make this problem prominent, it is not clear that this is indeed the case. The presentation of Section III.B could be improved, perhaps making some results Lemmas and the steps in the derivations a bit more clear. Figures 2,3,4 (the plots inside) are not legible in black and white. I don't agree that an SFR of 100m is reasonable. For most cases (except behind a sharp turn or an intersection) this is a LOS distance. Ensuring that at least nodes inside the SFR get the message might not be enough. Furthermore, these nodes might not be able to react in time, even if they get a relevant beacon, due to the distance. Perhaps approaches that simply try to supress redundant beacons altogether, rather than allowing all nodes to transmit to their neighbors but only within 100m, (i.e. space sampling or data fusion) might make more sense in very congested situations. Section IV.D on the location and power estimation is not that convincing. ======= Review 4 ======= *** Familiarity: Rate your familiarity with the topic of the paper. Expert (I conduct(ed) active research work in this topic) (4) *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). weak reject: top 50% of all submissions, but not top 30% (2) *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. The paper proposes a technique to reduce congestion in vehicular networks based on a combination of CS tuning, tx power control, and modifications to the backoff mechanisms. While some of the ideas in the paper might be potentially interesting, the proposed design, as well as the analysis used to motivate the whole approach, heavily build upon too many unrealistic assumptions about the radio propagation environment. *** Strengths: What are the major reasons to accept the paper? [Be brief.] The idea of combining CS range tuning, tx power control, and backoff modification to control congestion is potentially interesting. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] The effect of LOS/NLOS conditions on radio signal propagation is completely ignored. Analysis is weak and based on too many simplifying assumptions. Proposed approach heavily builds upon too many unrealistic assumptions about the radio propagation environment. *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. The main problem with this paper is that the authors do not adequately take into account the complexity of the radio propagation environment in vehicular settings. In particular, the effect of LOS/NLOS conditions, which has recently been recognized as very important based on on-the-field measurements (see, e.g., paper by Boban et al., "Impact of Vehicles as Obstacles in Vehicular Ad Hoc Networks", IEEE JSAC, Jan. 2011), is not considered at all in the paper. The authors ignore LOS/NLOS effects both in the analysis, and in the simulation-based evaluation. Based on recent measurement-based findings, the very same notion of CS range based on distance is not very meaningful in vehicular environments, not to talk about distance estimation techniques based on RSSI or similar techniques. Why not trying to used GPS information contained in beacons, when possible? When presenting the analysis, the authors say that they are not concern with MAC layer delay. However, MAC delay is a very important parameter in active safety applications, since very strict delay requirements must be fulfilled. Also, the analysis is based on many simplifying assumptions regarding not only the radio propagation environment (as commented above), but also the number of neighbors within CS and safety range, etc. In general, the plots reported in the figures are very difficult to read. The proposed technique heavily relies on the ability of quite accurately estimating vehicle positions based on radio signal parameters. This is extremely difficult to achieve in a vehicular setting, especially in a urban scenario. In accordance with this observation, the authors simulated only highway and rural scenarios. My guess is that the proposed approach would be not feasible at all in a urban scenario (and I am also not convinced that it is indeed applicable in highway scenarios, as commented above). In Table I, the P_min and P_max value are expressed in dB, instead of dBm. *********************************************************************************************************************************************************************************************** Comments for previous version submitted at INFOCOM 2012 *********************************************************************************************************************************************************************************************** ======= Review 1 ======= *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. In this paper the authors a MAC protocol for vehicular networks. The authors demonstrate the impact of the selection of an appropriate carrier sensing range on the collision probability. A CSMA based protocol is then proposed that uses a mix of various techniques such as power control. *** Strengths: What are the major reasons to accept the paper? [Be brief.] The impact of the carrier sensing range on collision probabilities is analyzed. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] There are a number of technical factors that have been ignored by the authors. The protocol's performance has not been thoroughly evaluated. *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. In this paper the authors motivate a different approach to controlling the congestion in vehicular networks. The authors propose changing the carrier sensing threshold of nodes, in conjunction with techniques such as power control and reverse backoff. The main criticism of the paper is that it fails to see the negative impacts of the proposed solutions. To start with, increasing the carrier sensing region may help with reducing the number of collisions. However, it will also decrease the spatial reuse and the overall network throughput. In the simulations section, the authors have not commented on the overall network performance and instead focused only on one metric: collision probability. Existing literature has explored the relation between the carrier sensing range and the interference range (which, in general, are not equal). It is well known that for ideal CSMA behavior, the carrier sensing region should be the same as the interference region. It would have been interesting to see how the authors' protocol performs against an approach where the carrier sensing range is set equal to the interference range. Another serious drawback of the proposed protocol is that it is dependent on the ability of nodes to determine the location of other nodes in its vicinity. The methodology proposed by the authors is not practical and also has flaws. Firstly, it is overly simplistic and ignores effects like fading which are common in vehicular environments. Also, the distance measured by the authors' methodology will result in an ambiguous location since a node at a distance X may be either in front of a given vehicle or behind it. The analysis and discussion on the collision probablities are based on many simplistic assumptions. The biggest of these is that the authors only consider one potential interferer for a given transmission (Section III.A). The analysis and the results will be quite different when one considers that interference power may be contributed by more than one node. There is a disconnect between the section where the authors describe the effect of carrier sense range on the collisions and the section where the protocol is proposed. The proposed protocol is a mix-and-match of many existing techniques such as power control and reverse backoff with one proposed by the authors. *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). C = (top 50% of reviewer's perception of all INFOCOM submissions, but not top 30%) - weak reject (2) ======= Review 2 ======= *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. Highlight the problem with 802.11p and propose a possible solutions based on controlling power, and rate. *** Strengths: What are the major reasons to accept the paper? [Be brief.] Highlight the observation that the cause of interference is different for nodes with different distances from the source and exploit this fact to improve delivery performance for vehicles closer to the source which are more important for safety reasons. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] The proposed algorithm depends heavily on the assumption that the relationship between received power and distance can be accurately modeled by the path loss model, which is likely to be not true in practice over short intervals in a mobile environment with loss of obstacles (vehicles). There is no experimental validation *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. This paper proposes changes to the IEEE 802.11p protocol to improve packet delivery for vehicles that are closer together. This is particular useful in a VANET environment where messages from nearby vehicles contain more urgent safety messages. The work relies on a number of observations/assumptions. First, for vehicles close by, interference due to hidden terminal is less. Second, it assumes the presence of capture effect. Hence, even with collision, packets with much stronger signal can be decoded. Third, a “perfect disk” radio propagation model whereby distance can be estimated (accurately) through power measurements and vice versa. Hence, the algorithm dynamically changes its behavior after a collision depending on how far the interferer is. While the proposal make sense given the assumptions, however, the assumption made on radio propagation makes the results strongly dependent on the assumed relationship between received power and distance . While distance is a dominant parameter, in practice, there is a lot of noise in the measurement and the estimation has lots of errors. Given that this is a safety application, there is a need to validate experimental results with some field measurements. Others: - There should be more details on the DSRC MAC behavior assumed, frame structure of time slots etc. - In Section III-B, it is mentioned that there is no exposed terminal problem on the vehicular control channel, why? - More details on the simulation environment, in particular, the radio model. *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). C = (top 50% of reviewer's perception of all INFOCOM submissions, but not top 30%) - weak reject (2) ======= Review 3 ======= *** Contributions: What are the major issues addressed in the paper? Do you consider them important? Comment on the novelty, creativity, impact, and technical depth in the paper. This paper proposes a new approach to improve MAC coordination in vanet scenarios mainly through manipulating carrier sensing mechanism. The idea is fairly interesting but it is hard to judge how novel it could be (in terms of its similarity to [11] and authors' own past works [4][10]. There are a number of technical concerns that authors might need to either improve or clarify. Also, the simulation is not conducted in a persuasive fashion. *** Strengths: What are the major reasons to accept the paper? [Be brief.] By manipulating carrier sense mechanism, the authors propose to enable vehicles to only collide with the nodes far away. As long as capture effect is taking its own role, the proposed mechanism (which is fairly aggressive) is able to accomodate more messages in VANET while minimizing the negative effect on safety messages for nearby vehicles. *** Weaknesses: What are the major reasons NOT to accept the paper? [Be brief.] - The novelity of this idea (comparing to [11] and [4][10]) is not clear; - The proposed mechanism seems to work well under the following scenarios -- linear highway topology, relatively simple fading scenario, and low to moderate vehicle density; It is not clear if the proposed idea is working for much more complex scenarios; - Many literature have demonstrated that the location mechanism by reverse engineering path loss model is not working in realistic situation; - the simulation is neither systematic nor persuasive *** Detailed Comments: Please provide detailed comments that will help the TPC assess the paper and help provide feedback to the authors. As summarized above, this paper has proposed an interesting idea which is worthy of thorough investigation, though the current shape of this paper is still not mature enough to be published in a major venue like INFOCOM. First, the authors need to do a better justification job how this paper is different from [11] and their own work [4][10]. Though authors stated that “it is first study proposing a new carrier sensing technique specifically designed for safety communication”, the reviewer admits that I fail to see how this claim could be supported. [11] is a fairly generic mechanism jointly adjusting carrier sensing and power control in ad hoc network, while this paper is targeted to VANET and safety communication – To this reviewer, it seems that this paper is a particular use case of [11] and a natural extension of [4]. If authors view differently, it would be beneficial for authors to clearly justify/clarify the major unique new contributions of this paper on top of [4][11]. Similarly, the content on reverse backoff mechanism (Section IV-E) seems to be directly borrowed from [10], even though this reviewer enjoys this controversial yet inspiring idea of RB. Second, the technical content of this paper needs to be further improved to become more rigorous. For example, (1) the designed protocol is highly customized for a linear road topology; as realistic VANET might cover simple linear topology as well as much more complex situations like intersection or overlay highways; (2) the fading effect is fairly simple and manageable, so that the vehicle V could estimate what level of interference imposed on station S situated at the border safety range of vehicle W; again, that might be another luxury assumption which is not valid; (3) density used in this study (43 veh/lane/parameter) is not as congested as possible (like a bumper-to-bumper traffic jam in Paris). Corresponding to these issues, the validation/simulation process also does not consider these complex and realistic scenarios, so the validation process is not as systematic as the reviewer expects, which is not convincing enough to readers. The paper is reasonably well written, while further improvement in terms of organization and writing could be expected. The reference is adequate. There are a number of local comments: 1. Page 3, Sec III-A, at the end, why “exposed terminal issue does not exist in VANET”? 2. Page 3, sec III-B, at the beginning, you mention “multiple radio propagation problem”, do you mean “multiple path propagation problem” or you mean “multiple radio access technologies co-existing problem”? 3. Page 5, Sec IV-A, your protocol’s design is fairly aggressive, essentially it tries its best not to backoff/wait as long as it will not endanger the safety border’s safety margin. As you said in the paper, it will increase the concurrent transmission message at a given time but at the cost of endangering the faraway nodes. This reviewer can see such an aggressive behavior could benefit at a given region of parameter space, but I doubt this aggressive behavior could even degrade the performance at other region of parameter space. Could you shed light on this concern? 4. Page 6, Sec IV-D, many literature have demonstrated that the location mechanism by reverse engineering path loss model is not working in realistic situation; 5. Page 6, sec IV-D, why you mention “for vehicle safety message, the location of transmitter is already in MAC header”? GPS location of vehicle is in safety message, which is a network-layer or even app-layer message. 6. Changing standard is not a trivial task, even though it might be as small as add a few bits in the headers. Backward compatibility issue is always very difficult in standard community. *** Recommendation: Your overall rating (Please try giving as few borderlines as possible). C = (top 50% of reviewer's perception of all INFOCOM submissions, but not top 30%) - weak reject (2)