Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a well-executed study on communication scheduling for bandwidth-limited multi-robot exploration. The authors identify and characterize a phenomenon they term the "synchrony tax"—the efficiency penalty incurred when robots broadcast on a common schedule, causing their shared beliefs to go stale simultaneously and leading to redundant exploration. The work combines theoretical analysis, extensive simulation experiments, and a hardware demonstration to support its claims. It should be improved by addressing below comments:
1. The proof of Proposition 2 is not provided in the main text (only in Appendix A), and the bound itself is relatively weak. The statement "where wij(t)=K[∥pi−pj∥≤2R+δ] marks contending pairs and Φ is non-decreasing" is underspecified. What is the form of Φ? Is it a known function or an abstract envelope? The practical utility of this bound for predicting actual overlap values is unclear.
2. The proof of Proposition 3 assumes the symmetric, fully-contending case with equal outgoing weights  ak≡a. However, the statement claims this minimizes the overlap envelope for the general case. The transition from the restricted case to the general case needs justification. The finite exhaustive search (for  N≤6,T≤16) is acknowledged but is insufficient as a general proof.
3. The proof of Theorem 1 relies on referencing [11] for the key optimality result. While this is acceptable, the theorem's novelty relative to existing work is not clearly articulated. The contribution appears to be the application of this scheduling framework to the multi-robot exploration domain with task-specific weights, but the theoretical novelty should be stated more explicitly.
4. The decisive "phase-only control" compares staggered vs. synchronized periodic schedules while allowing byte volumes to differ (staggered transmits up to 16% more bytes). While the authors argue this "sharpens the conclusion," it actually introduces a confound. The byte-matched comparison in Table 1 uses different schedulers (Random, Event-triggered) rather than the same schedulers with controlled byte volume. A more rigorous design would either (a) implement byte-identical staggered vs. synchronized schedules, or (b) use byte volume as a covariate in all comparisons.
5. The paper reports many p-values with Holm correction within each sweep but does not clearly indicate which comparisons were pre-specified versus exploratory. The statement that "Our conclusions rest on the headline effects, which remain significant under Holm correction within each sweep, not on individual marginal points" is somewhat defensive. The presence of marginal effects (e.g., the 20% packet loss dip, the 40m range marginal point) suggests the effect may not be as robust as claimed.
6. Only one paired trial per condition is reported for the hardware validation (Section 4.9). This is insufficient to draw statistical conclusions. The authors state "one paired trial per condition" but then report values like "34% reduction" as if these are statistically meaningful. The hardware results should be presented as a qualitative demonstration or feasibility proof, not as quantitative evidence.
7. The paper focuses heavily on the DARS/Ariadne line of work in Multi-robot exploration literature but does not sufficiently engage with other communication-aware exploration approaches, such as:
o Decentralized frontier assignment with communication constraints
o Information-theoretic approaches to multi-robot exploration
o Hierarchical or hybrid planning approaches that handle communication constraints
o Other contraints for individual robots in exploration tasks, such as path planning, locomotion control, should be cited and discussed, such as:
[1] L. Chen et al., "Optimized Foothold Planning and Posture Searching for Energy-Efficient Quadruped Locomotion over Challenging Terrains," 2020 IEEE International Conference on Robotics and Automation (ICRA), Paris, France, 2020, pp. 399-405, doi: 10.1109 ICRA40945.2020.9197135.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript focuses bandwidth-constrained multi-robot exploration and studies the impact of communication scheduling (rather than just communication frequency) on team coordination. The proposed de-synchronised scheduling strategy is well-motivated and augmented by theoretical analysis, simulation studies and hardware experiments. Overall, the manuscript is a well-structured and technically sound document; however, some sections are excessively long and the presentation of the theoretical development can be improved for clarity. Furthermore, providing a clearer discussion of the practical limitations, scalability of the method in more adverse network situations, and computational considerations for larger teams would further strengthen the paper. In conclusion, this work contributes considerably to the existing literature, though some minor clarifications and refinements are necessary before publication.

However, the following elements need to be addressed in the manuscript:

• An explicit novelty statement is missing from the Introduction statement. A clear and concise novelty statement can help the readers understand the novel element of this research study.
• The main assertion points to an improvement in team performance being due to a decrease in the amount of time due to the stale data at the time of data being processed rather than the overall amount of time the stale data is around the team. What evidence based on appropriate experiments points to peak-level stale data as the key cause & further supporting evidence, or greater occurrence between peak-level or average level of time elapsed before data was received, should be conducted to make this conclusion substantially stronger?
• The theoretical premise of the efficiency of exploration is primarily from the group's stale shared beliefs about the time data is received at their specific location. Under what circumstances will this statement hold true; for example, have little deviation in localizing between where data should have been and where it actually has been, experiencing continuously changing objects where a team must keep track of, inconsistencies between a team’s representation of where and what data was received and finally communication delays greater than the time required to complete the task? Defining the area of where the statements about exploratory behavior are valid will make the theoretical assumptions much clearer.
• The proposed framework was evaluated using a learned framework as a basis for communicating within a group of explorers. To what extent were the improvements realized due to the specific tasks and/or whether they would have been realized without any changes of direction due to other teams attempting similar tasks? Validating this premise with very different ways of working together as a group will enhance the confidence in the applicability of the improvements.
• OW-Desync was motivated by theory for heterogeneous interference graphs, but experimental results have shown that it mostly performs as well as (not better than) the simple form of even staggering in the exploration task. What types of real world scenarios will create a measurable benefit from using OW-Desync and what level of network heterogeneity is necessary to justify its added complexity?
• The only change made to the communication schedule is the timing of the transmission; the exploration policy remains the same. Will there be a chance that long-term phase alignment could occur naturally after an extended period of operation due to asynchronous execution, variable computation time and network jitter, and thus negate some of the benefits associated with de-synchronising? Having quantitative data regarding the stability of the scheduling mechanism over a long period of time would be beneficial in determining whether or not the mechanism is robust enough for use in practice.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript studies the effect of communication scheduling on bandwidth limited multi robot exploration. The authors identify a synchronization problem in which robots broadcast at the same time and consequently use stale shared information at similar periods. A desynchronized communication scheduling approach is proposed to distribute robot transmissions over time while maintaining the same nominal broadcast rate. The simulation results show that the proposed scheduling approach can reduce travel distance and sensing overlap compared with synchronized communication. The authors also present a staleness based analysis to explain the observed behavior and extend the scheduling approach to heterogeneous and range limited communication using OW-Desync. In addition, a four robot experiment using a real WiFi network is presented to evaluate the effect of communication timing on team staleness. The topic is interesting and the results are promising. However, several claims concerning the mechanism and hardware validation are stronger than the experimental evidence presented in the manuscript.

Strengths
1. The manuscript addresses an interesting communication scheduling problem in bandwidth limited multi robot exploration.
2. The comparison between synchronized and desynchronized communication is clearly designed and evaluated under matched nominal broadcast rates.
3. The study combines simulation, theoretical analysis, and a real WiFi experiment to investigate the proposed communication scheduling concept.

Comments and Suggestions
1. The hardware validation does not fully validate the proposed exploration mechanism. The four robot experiment evaluates communication timing and team staleness. However, sensing overlap, travel distance, or other exploration performance metrics are not measured in the hardware experiment. Therefore, the experiment only validates the relationship between communication scheduling and peak team staleness. It does not validate the complete relationship from communication synchronization to staleness, redundant sensing, and exploration performance. The scope of the hardware validation should be clearly stated.
2. Several statements in the Abstract, Introduction, and Conclusion overstate the hardware results. The manuscript states that the mechanism or task level gain is reproduced on four robots over real WiFi. However, no task level exploration results are reported from the real robot experiment. These statements are not fully supported by the presented results and should be revised. The authors should clearly state that the hardware experiment validates the communication timing to staleness component of the proposed mechanism only.
3. The causal claim concerning peak team staleness is too strong. The manuscript states that a higher peak team staleness is the cause of redundant sensing and increased travel. The reported correlation between peak staleness and sensing overlap and the negative control experiments provide useful evidence supporting this explanation. However, the theoretical relationship is derived under the deconfliction assumption, and the experimental results do not directly establish causality. The claim should therefore be stated as a supported mechanism or a primary contributing factor rather than a confirmed cause.
4. The practical exploration benefit should be experimentally verified on the real robotic system. Since the main contribution of the manuscript concerns multi robot exploration efficiency, a real robot experiment comparing synchronized and desynchronized communication in terms of travel distance and sensing overlap would significantly strengthen the work. At present, the real WiFi experiment verifies communication freshness but does not demonstrate that the reported simulation level exploration benefit can be reproduced in a physical robotic system.
5. The limitations of the proposed mechanism should be discussed more clearly. The results show that the benefit of desynchronization depends on how strongly the exploration policy uses the shared teammate belief. The benefit becomes weak or disappears for planners with limited belief based deconfliction, and a small performance cost is observed for the myopic planner. This is an important limitation of the proposed approach and should be clearly stated in the Abstract and Conclusion rather than presented mainly as supporting evidence for the proposed mechanism.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed the comments I proposed. The manuscript can be accepted now. 

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have addressed all the comments.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have adequately addressed all of my previous comments and appropriately revised the manuscript. I have no further questions or comments.