======= Review 1 ======= > *** Recommendation: Your overall rating. Weak Accept (Top 20%) (4) > *** Contributions: What are the major issues addressed in the paper? Do you consider them important? (Pls. comment explicitly on the relevance of the paper to MASS topics, the technical depth and the importance of the problem addressed.) [Be brief] This paper studies the impact of the photovoltaic energy harvesting paradigm on the duty cycle of sensor nodes through both indoor and outdoor datasets analysis. Some findings are observed. Firstly, very high duty cycles, in the order of tens of percents, can be achieved even using the static duty cycle approach in outdoor scenarios. As for the indoor scenarios, since the energy harvesting opportunities are usually changing, the dynamic duty cycle is exploited. Additionally, the author designs a dynamic duty cycle strategy by introducing battery level variation to achieve better overall activity. This is the novelty of this paper. But there are still some places need to be taken cared. > *** Strengths: What are the major reasons to accept the paper? [Be brief] 1, This paper experimentally studies the impact of the photovoltaic energy harvesting paradigm on the duty cycle of sensor nodes in the wireless sensor networks, which shows some light on the design of sensor networks with energy harvesting. 2, Through data set analysis, relatively high duty cycles (more than 30%) for sensor nodes is feasible in outdoor scenarios, even when a static duty cycle approach is used. 3, In the indoor case, the analysis shows that the dynamic duty cycle approach based solely on the battery residual energy does not necessarily achieve better results than the static approach. 4. Through choosing optimized parameters, the author finds the dynamic duty cycle approach by adding supplementary information regarding the battery level variation to the duty cycle computation becomes more efficient than the static one. > *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief] 1, The author gives a definition of continuous operation and use it as the metric to verify the performance of different strategies. However, the description of the continuous operation is un-rigorous, which would make the author confused. 2, In the dynamic duty cycle approaches, this paper tries to adjust the duty cycle to achieve better overall activity. However, when a sensor node changes its duty cycles, such as increase or decrease the duty cycle, it needs to notify its neighbors to enable communication. This would lead to additional energy consumption. How to do with such cases? The author should also take it into consideration in the experimental studies. 3, In the section of static duty cycle analysis, the metric of DCmax is compared and gives an equation to compute it. The equation is a little confused when c.o. denotes whether a time slot can be continuous operation. 4, In this paper, the author only gives two metrics, i.e. the continuous operation and overall activity to evaluate different duty cycle strategies. Is there any other metrics used by other literatures to evaluate the performance > *** Detailed comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper, as well as feedback to the authors. 1, There are too many symbols used in this paper, the author had better list a table of used symbols in this paper. 2, In the subsection of III.A, the equation of calculating the harvested energy Eh(t), the function of E[] may be confused when it is used together with other symbols. 3. In the Fig.4, there is no introduction to “BREV-r” in this subsection. The author should give a description before using it. ======= Review 2 ======= > *** Recommendation: Your overall rating. Borderline (top 30%) (3) > *** Contributions: What are the major issues addressed in the paper? Do you consider them important? (Pls. comment explicitly on the relevance of the paper to MASS topics, the technical depth and the importance of the problem addressed.) [Be brief] The paper presents a model for duty cycling sensor networks with energy harvesting. The authors examine both outdoor and indoor scenarios and develop an adaptive duty cycling/power management algorithm that takes into account battery variation (real-time readings of) to adapt application duty cycle to the net gain or loss energy in the system. The result is that the applications can achieve optimal use of the harvested energy, and improve the overall duty cycle of the application across a range of operating conditions. > *** Strengths: What are the major reasons to accept the paper? [Be brief] Adaptive duty-cycling of applications based on energy harvesting is an under-researched topic, with authors proposing a solution that improves prior work. > *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief] The work lacks details on the size of the solar panels used in outdoor/indoor datasets, so it's hard to judge the relevance to WSNs. The model assumes a simple sensing model, it would be nice to incorporate event-based sampling rather than only static DC sampling. > *** Detailed comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper, as well as feedback to the authors. The authors correctly observe that static duty cycles are often unachievable and sub-otpimal in systems where energy is dynamically harvested during the system operation. The BREV approach they present shows an improvement over the static DC across a range of simulated scenarios, while providing an operational system in all scenarios. The limitation of this work is the lack of empirical experimentation - it is based on simulations that use empirically measured harvested energy and an idealized model of the wireless sensor. It was unclear, what size of solar panel did the authors assume - specifically the los angeles data set seems to harvest close to 900J per hour which indicates a fairly large solar panel compared to the one used in the indoor new york deployment. Some other assumptions (such as radio energy modeling) are also outdated, especially for long range outdoor applications - transmission power can be orders of magnitude larger than reception power to allow for long range. Finally, while the paper has shown that the BREV algorithm is not too sensitive to r, it hasn't shown the same for DC_MAV - the value of this parameter will have significant impact on the performance. Is it possible to infer this parameter to optimize the algorithm performance? One approach to setting/adapting the DC_MAV would be to use the expected harvested energy for the day (there's been some work in WSNs for energy harvesting prediction). I suggest the authors to look into this in their future work. ======= Review 3 ======= > *** Recommendation: Your overall rating. Borderline (top 30%) (3) > *** Contributions: What are the major issues addressed in the paper? Do you consider them important? (Pls. comment explicitly on the relevance of the paper to MASS topics, the technical depth and the importance of the problem addressed.) [Be brief] The authors propose a detailed energy harvesting analytical model for a node, considering the energy collection, energy management and energy consumption. The topic seems interesting. > *** Strengths: What are the major reasons to accept the paper? [Be brief] The topic seems novel and interesting. The conclusion of indoor and outdoor scenarios is useful. > *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief] Current paper needs some efforts to improve the writing. Many sentences have grammar problems, e.g. wrong tense. Please pay attention to make the paper more readable. More experiments should be added to make the paper convincing. > *** Detailed comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper, as well as feedback to the authors. 1.The English written need to be improved. For example, in the abstract section, the sentence We show that, for thestatic duty cycle approach in outdoor scenarios, very high dutycycles, in the order of tens of percents, are achieved, eliminating the need for additional energy conservation schemes looks quite strange. Similar kinds of sentences are widespread in the paper. Please double check the English written to make it more readable. 2.The authors claim that the approach always outperforms static solutions, no matter whether the perfect knowledge of harvestable energy is assumed. However, there is not enough proof in the experimental part, the bars in Fig. 3 only compared the static solution with the energy harvesting. Seems more experiments should be added. 3.The authors said there are other prediction-free algorithms like Vigorito et al. and Yoo, why not compare the experimental results with them in the experimental parts. ======= Review 4 ======= > *** Recommendation: Your overall rating. Borderline (top 30%) (3) > *** Contributions: What are the major issues addressed in the paper? Do you consider them important? (Pls. comment explicitly on the relevance of the paper to MASS topics, the technical depth and the importance of the problem addressed.) [Be brief] The paper proposes a new approach for computation of duty-cycles in solar-powered energy harvesting network by integrating the variations of battery levels. The authors provide comparison with static and dynamic approaches when the perfect knowledge of the harvestable energy is given. Overall the paper makes a simple contribution of adding variance of battery levels to formulation of duty-cycle. While this is an interesting approach, it is not sufficiently novel (temporal variations in charging times as well as harvesting rates have been implemented for past years by the community, and the battery level variations are directly related to these). Besides, there are problems in the evaluation of the BREV approach. > *** Strengths: What are the major reasons to accept the paper? [Be brief] The paper is well written and readable, and has a good tutorial value in establishing the background. There are ample references for the reader to pursue further knowledge on the subject. The proposed idea is described well and explained clearly. > *** Weaknesses: What are the most important reasons NOT to accept the paper? [Be brief] The paper lacks sufficient novelty and depth of exploration. Also, the paper lacks experimental results to show its practically. Duty-cycle based networks have been implemented and used for many years, and providing experiments for the proposed approach would be a necessary part. There are no analytical proofs that this system is indeed better and can converge. The simulation results are for an extremely limited set of scenarios, and there is no comparative evaluation with any other work. > *** Detailed comments: Please provide detailed comments that will be helpful to the TPC for assessing the paper, as well as feedback to the authors. While the idea of integrating battery level variations to compute duty-cycles is interesting, this idea is not novel and very similar to integrating temporal variations of charging time and temporal variations of harvesting rates. There are many papers that authors have not cited and cover these ideas such as: “Device Characterization and Cross-layer Protocol Design for RF Energy Harvesting Sensors," Elsevier Pervasive and Mobile Computing Journal, vol. 9, no. 1, pp. 120-131, February 2013. The performance evaluation is rather weak. The authors have not provided any experimental results for this approach. The most related papers in the literature that introduce a new duty-cycle approach has the real experiments part and are not complex to implement. Finally, one of important and effective parameter in the performance of the proposed approach is the size of battery that can mitigate the variations in the harvesting energy. However, the authors have not explored BREV for various battery sizes in indoor scenarios. #################################################################################################################### ########################### Reviewer comments for previous version submitted at SECON 2016 ####################### #################################################################################################################### =========================================================================== SECON 2016 Review #72A --------------------------------------------------------------------------- Paper #72: Energy Harvesting Wireless Sensor Networks: From Characterization to Duty Cycle Dimensioning --------------------------------------------------------------------------- Overall merit: 2. Weak reject Reviewer expertise: 4. Expert ===== Paper summary ===== Paper examines approaches to duty cycling in energy harvesting networks. It demonstrates that duty cycling approaches often proposed for energy harvesting networks (e.g., those based on node battery levels) are inadequate, and presents an approach that is based on battery level dynamics that addresses the shortcomings of other techniques. ===== Strengths ===== Paper makes valid, reasonable, important observations. For example, it is definitely true that battery level- based approaches are inadequate for energy harvesting networks, and definitely true that one can do better with the approaches based on battery variability. Paper examines in some depth node performance achievable in different light energy conditions with different energy storage types. ===== Weaknesses ===== Paper's contributions, while reasonable and valid, are not sufficiently developed to warrant being included in the IEEE SECON program. In this reviewer's opinion, this paper, as-is, would make for a nice self-contained publication in a small workshop of a major conference. Examples of extensions of the work that would bring it closer to IEEE SECON level of a contribution: - In-depth examination of performance of BREV, analytically (are there performance bounds that can be proven about it?) - Examining BREV experimentally, via implementing it in real-world devices (is BREV practical in real systems, where battery dynamics may or may not be easy to track closely?). One concern about paper's results: I believe authors' insights into performance of outdoor systems are influenced by their choice of a location to examine (Los Angeles). Observed node behavior in outdoor conditions is likely to be very different for locations with more cloudy days and more weather variety in general (e.g., in Northeast locations). Moreover, what about shading? - in practical scenarios, unless an energy harvesting node is under direct unobstructed sunlight, its' energy situation will be more challenging than one would expect from the US Department of Energy Data. Overall, I believe that author's conclusion on node duty cycle in outdoor deployments (V.A) may be valid under specific circumstances, but is otherwise overly optimistic. =========================================================================== SECON 2016 Review #72B --------------------------------------------------------------------------- Paper #72: Energy Harvesting Wireless Sensor Networks: From Characterization to Duty Cycle Dimensioning --------------------------------------------------------------------------- Overall merit: 3. Weak accept Reviewer expertise: 4. Expert ===== Paper summary ===== This paper studies energy harvesting powered wireless sensor nodes which conserve energy through the use of duty cycles. They present an energy harvesting model which incorporates energy harvesting, energy storage, and energy consumption. Based on the model, they consider applying duty cycles to the nodes operating in either indoor or outdoor scenarios. Both static and dynamic duty cycles are considered. Using real world irradiance data sets from [21], [40], the authors show that static duty cycles are sufficient in outdoor locations due to significant amounts of solar energy. In indoor scenarios, the authors show that dynamic duty cycles do not necessarily outperform static duty cycles. Thus, they propose their own dynamic duty cycle protocol, BREV, whose performance always outperforms static duty cycle protocols. ===== Strengths ===== - The authors proposed a detailed, yet simple, model of energy harvesting, energy storage, and energy consumption. - The authors evaluate a state of the art duty cycling protocol (i.e., BRE) using real datasets in both indoor and outdoor locations, demonstrating strengths and weaknesses. - The authors propose and evaluate BREV, and show that it resolves the inefficiencies of BRE. - The paper is clear and well written. ===== Weaknesses ===== I have numerous reservations regarding some of the assumptions of the paper which undermine some of the main conclusions. Primarily, the slot size in this work is fixed at 1 hour. I question whether their main result, that BRE suffers from several inefficiencies, is simply a byproduct of this. Why was 1 hour time slot chosen? How would BRE perform is a time slot of 1 second was chosen -- it would seem to me that it would be able to adapt to initial battery depletion on a much faster scale. Justification of this, and other, assumptions would significantly improve the quality of the paper. For more examples, see detailed comments below. Detailed comments: * Assumptions: Some of these make for a clean model, yet it would be nice if they were discussed in detail and how such conclusions would behave in practice (perhaps through an experimental evaluation). P2, 2nd column: the cell efficiency \eta_{PV} and battery discharge efficiency, are assumed to be a constant. In many scenarios this is not true. Cell efficiency varies with degradation over time, temperature, and operating voltage. Indeed, this is difficult to capture in modeling. P2, 2nd column: I_{gh}(t), the irradiance for a single slot, is listed as constant for a slot size. This should be representing the _average_ irradiance in a slot. Depending on the slot size (in this case 1 hr), the irradiance will likely vary (in some cases, significantly). P2, 2nd column: CPU power consumption is listed as negligible. In general, this is a safe assumption. However, for very small duty cycles considered for indoor scenarios (e.g., ~1%), if the CPU is left on, even in a low-power mode, it will be powered on for approximately 100x the length of the radio, and thus may not be negligible. P3, 1st column: The average power consumption I_{avg} of the radio is aggregated for the transmit, listen, and receive mode for a radio. In practice, is it true that the device spends equal time in each of these states? It would seem that, depending how the duty cycle is provisioned, a node might spend additional time in the listen state than in the receive state, for example. * The main performance metric considered is “overall activity”. This metric is not well explained. From (8), it seems that it is just the average on time per slot achieved over the period \tau * Section IV: Why consider one average day as representative for each month? This significantly reduces the variability in the energy harvested over time. Given that the data is available, it would be better to use real data for energy harvested in each day. * As described in section VI, the dynamic approach is superfluous in outdoor scenarios. Indeed, the outdoor scenario is not all that interesting (due to power harvesting up to 1000x higher). The interesting parts of this paper focus on indoor scenarios. * The BREV protocol has control theory undertones. For example, BRE is a proportional controller and BREV is a proportional + derivative controller. This is similar/should be compared/contrasted from other work in this area (e.g., T. N. Le et al. Power Manager with PID controller in Energy Harvesting Wireless Sensor Networks. Proc. IEEE GreenCom’12) * Minor typo: Last paragraph of Section IV, first sentence: “Remaining” -> “remainder” =========================================================================== SECON 2016 Review #72C --------------------------------------------------------------------------- Paper #72: Energy Harvesting Wireless Sensor Networks: From Characterization to Duty Cycle Dimensioning --------------------------------------------------------------------------- Overall merit: 1. Reject Reviewer expertise: 4. Expert ===== Paper summary ===== This paper studies duty cycle dimensioning in PV energy harvesting wireless sensor networks in both outdoor and indoor environment. It proposes a BREV which outperforms static duty cycle approaches. ===== Strengths ===== The paper uses measured PV data to evaluate the performance of the duty cycled wireless in both outdoor and indoor environment. A dynamic duty cycle approach that considers both battery residual energy and battery level variation is proposed for better performance. ===== Weaknesses ===== 1. The title of this paper is not appropriate. It implies study of general energy harvesting wireless sensor networks, but the paper only discussed typical PV energy harvesting whose profile is very different from other energy sources. The results of this paper may not apply to other energy harvesting wireless sensor networks. 2. The paper uses datasets from NREL to model the input of the outdoor energy harvesting. However, according to Fig. 1(a), it only uses the data of sunny days during which data curves are parabola shaped. During the days when sunlight is block, the available energy reduces and may vary significantly. Due to such uncertainty of energy availability, DC_{max} cannot be guaranteed. Given the situation, it is not safe to propose the first statement in Section I: "Operation with relatively high duty cycles (more than 30%) is feasible in outdoor scenarios, even when a static duty cycle approach is used. This questions the need for sophisticated energy management techniques in the case of outdoor WSN." 3. The paper cites [8] to illustrate the high duty cycle that can be achieved by energy harvesting sensors. However, [8] focuses on the design of a signle sensor node, and the performance is evaluated on a single mote. In contrary, this paper studies sensor networks, which usually consist of hundreds or thousands of sensors, and the energy consumption increases dramatically due to the multihop data transmission from sensors to base stations. A sensor that can individually work at 20% duty cycle may not be able to achieve such a high duty cycle in a network. Unfortunately, this paper doesn't realize that such significant difference takes the factor into account, thus the conclusion is not convincing. #################################################################################################################### ########################## Reviewer comments for previous version submitted at Infocom 2016 ###################### #################################################################################################################### ======= Review 1 ======= > *** Paper Summary: Please summarize the paper in your own words. This paper tries to improve the duty cycle performance with harvested energy. By analyzing the energy harvested at different locations, the authors claim that the outdoor scenario is different from the indoor scenario, calling for a dynamic duty cycle approach. Compared with the static duty cycle ratio, this approach has better performance without the knowledge of future harvest able energy. > *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. The authors analyze the harvestable energy at different locations using public traces. They also introduce one dynamic duty cycle approach based on residual battery energy and current energy changes. Simulation result demonstrate that the new dynamic approach has better performance. > *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. "However, this paper has several serious drawbacks. 1) This paper relies on a few unrealistic assumptions, for example, it uses one charge/discharge efficiency parameter to characterize the battery behavior and uses the fixed parameters to describe the power consumption on the wireless sensors. They are too theoretical for designing energy efficiency systems, and are unable to describe sensors' behaviors in the real scenarios. 2) The authors claim that predicting battery level is not feasible, so they need to conduct detailed analysis on why this is true. This paper currently lacks convincing reasons for this claim. 3)This paper is based on public traces only, without the validation from real testbeds. In this reviewer's opinion, a real testbed enables the authors to conduct a throughout analysis on the collected energy given different environments. 4)The authors claim that the DCVR algorithm outperforms the DSR algorithm, which only considers about the residual energy. However, the performance gain is little for 4 indoor cases, and the performance comparison is missing in the practical case evaluation. Overall, DCVR does not have a clear advantage when compared with DSR algorithm. 5) The paper only compares its DCVR algorithm with DSR algorithm, without a detailed performance evaluation for other existing works. For example, Adaptive Control of Duty Cycling in Energy-Harvesting Wireless Sensor Networks, is also a work which changes the duty cycle without the prior knowledge of the environments. The authors need to conduct a detailed analysis about the difference between this algorithm and other algorithms in the same field. 6) This paper is not well written. It is difficult to understand why the authors introduce some parameters, such as DCmav, and they do not give concrete and specific reasons to convince the readers. > *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. This paper is not well written. It is difficult to understand why the authors introduce some parameters, such as DCmav, and they do not give concrete and specific reasons to convince the readers. > *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. ======= Review 2 ======= > *** Paper Summary: Please summarize the paper in your own words. This paper compares static and dynamic duty cycle approaches in energy harvesting wireless sensor networks. It proposes a dyncamic duty cycle mechanism that considers both the battery residual energy and the charge/discharge behavior of the battery. > *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. The problem considered in this paper is important and interesting since energy harvesting capabilities bring new challenges into the system design. > *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. The contribution of the paper is marginal. It is not surprising that dynamic duty cycle outperforms static duty cycles, and dynamic duty cycle with more control signal inputs outperforms the one with less inputs. The paper does not show how the parameters used in the proposed approach is determined. For example, the residual energy threshold E_th and the weight parameter r. > *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The notations used in the paper can be improved. For example, "max" and "MAX" are both used. > *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank. Are there existing work in dynamic duty cycle that considers both the battery residual energy and the charge/discharge behavior of the battery? ======= Review 3 ======= > *** Paper Summary: Please summarize the paper in your own words. The authors propose an energy harvesting model for solar powered wireless sensor nodes that combines the collection of energy, its consumption and then management. From this they produce a duty cycling dimensioning scheme. Using data sets of energy opportunities they illustrate that the scheme can remain sustainable, with quite high duty-cycles. In indoor situations they show that static policies do not perform any worse than adaptive ones. So they took battery level variation to inform an adaptive duty cycle, and show that the node is active longer; however this approach expects full perfect knowledge of harvestable energy. > *** Strengths: What are the main reasons to accept the paper? You may comment on the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. Energy harvesting and energy management for sensor networking is a highly important and timely piece of work. Their result shows that a relatively simple adaptive energy management system can increase sensor node lifetime. > *** Weaknesses: What are the main reasons NOT to accept the paper? Again, think about the importance of the problems addressed, the novelty of the proposed solutions, the technical depth, potential impact, and the writing. Your overall rating should be supported by your review. Adaptive duty cycling for wireless sensor networks has been around for some time and much of this body of work is not reflected in the related work/background section. As far back as 2006, Adaptive Duty Cycling for Energy Harvesting Systems by Hsu et al in Low Power Electronics and Design, ISLPED'06 examined residual battery as an indicator for the power management system. Their adaptation to take battery state behaviour into account is very incremental. The authors have to tighten up and clarify some of their claims that justify their approach. One example is ' indoor PV energy is not periodic and hardly predictable'. Well this depends on the indoor environment, a home perhaps not, a factory or office space it is much more. Their fig 1b also shows this. Also stating in sensor nodes the transceiver out weighs the other hardware modules...again yet for little sensor, but many serious or industrial sensors this is not the case. It is unclear what the harvesting opportunity across the space is...do they assume homogeneous or heterogeneous (like in a park with trees and shadows etc) and in terms of spatiotemporal dimensions. In general the results as presented are rather obvious. Is overall activity a better metrics to understand behaviours than packet loss, delays etc? > *** Quality of Writing: What is the presentation quality of this paper? A paper not well-written is not good for INFOCOM reputation and will have difficulty in attracting citations. Your overall rating should take this into consideration. The writing is mostly reasonable with a few small mistakes.The layout flows well. Graphs and diagrams are of a reasonable size. Well, except figure 8 could have been made more clear. > *** Additional Comments: Additional comments (if any) that you would like to provide to the authors. Please do not repeat what you stated above. If none, leave the following blank.