An Improved Ant Colony Algorithm for Urban Bus Network Optimization Based on Existing Bus Routes

Round 1

Reviewer 1 Report

This manuscript aims to reoptimize the bus network based upon the existing bus routes. Further, an ant colony algorithm is proposed to solve this interesting issue. Results indicate that the proposed method is available to generate preliminary route networks for new urban areas without bus line coverage and also provide a reference solution for the adjustment in areas with uneven spatial coverage of bus lines. Overall, Overall, the paper is well written and its logic is clear to me. The reviewer has only a few questions or concerns:

In Figure 3, the figure resolution is low, and please update the figure with higher resolution. Moreover, the bus route [(181, 173), (173, 264),…] should be illustrated more clearly. For each figure, legend is necessary for the readability (for example, what’s the meaning of the blue line as shown in Figure 3; what’s the meaning of dashed red line in Figure 4).

In my opinion, a model should be designed to represent the detailed process of optimizing the bus network. The authors only state the elements of the network topology, but do not introduce how to formulate the model. For instance, how to formulate the existing routes, and what are the decision variables and necessary constraints when reoptimize the bus network. Later, the ant colony algorithm can be employed to solve this model.

Some relevant references seem to be missing, which provide beneficial findings of bus route/network design by using big-data techniques or optimization methods:

A partial-fréchet-distance-based framework for bus route identification. IEEE Transactions on Intelligent Transportation Systems, 2021.

Short-term forecasts on individual accessibility in bus system based on neural network model. Journal of Transport Geography, 2021, 93, 103075.

A two-phase optimization model for the demand-responsive customized bus network design. Transportation Research Part C: Emerging Technologies, 2020, 111, 1-21.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The article is  devoted to the optimization of existing passenger transport routes in the city. An algorithm based on the analysis of passenger flows is used, which is absolutely justified for solving this problem. The object and subject of the study are clearly defined. The presented methodology and the mathematical operators used correspond to the achievement of the goal. To the authors I have the following remarks:

1. The work does not take into account the parameter of accessibility of stopping points of the passenger transport network. How is the network coverage area of the city analyzed in the system?
2.  Also, the issue of the density of the passenger transport network requires coverage. The paper presents an indicator of the proportion of bus lines to the length of urban roads.
3. Have the authors compared the obtained values with any normative values of density and accessibility?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors present an application of the ant colony optimization heuristics (ACO) for the optimization of bus lines. The optimization of the routes is presented using the example of a section of the bus network in the Chinese city of Zhengzhou.

Generally, the article is well structured, well written, features several appropriate maps, tables and figures for the presentation of the methodology, the study area and the results, and contains a decent literature review. The methodological approach is sound and well presented, the conclusions are overall valid and interesting.

In detail, though, there are several aspects that deserve some improvement prior to publishing.

• First, the literature review lacks some relevant articles, in terms of important basic contributions, existing work in the area of application of the optimization of public transport lines and the specific variants of ACO implementations that are applied here. To name just a couple aspects:
  • The authors refer quite often to Dorigos work; I wonder why there is no reference to well-known contributions such as the Phd-Thesis by Blum 2004, Cordon et al. 2002 or Dorigo and Stützle 2004 – instead some quite unknown and uncited Chinese references are given.
  • A quick look at Google Scholar also indicates numerous recent applications of ACO to the case of bus route planning, for example by Calabrò et al. 2020, Hu et al. 2019, Yu et al. 2005, Vitis and Axhausen 2009, Huo et al. 2014, all of which are not listed. Here, a more specific delimitation of the presented contribution would be desirable.
  • On the other hand, the relevance of citing the literature in the first section on pace 2 is not apparent to me.
• On page 1, the authors refer to an abstract increase in scientific aspects and convenience being targeted by the usage of ACO. I wonder what this is referring to. Indeed, planning is often done by domain experts, that to this day often rely on individual experience rather than mathematical models. Is this what the authors refer to?
• On page 2, the authors claim that most alternative approaches would consider random lines as alternative solutions. I would suppose that this depends on the way the network representation as graph is chosen. The argument is repeated on page 11.
• On page 1, the authors refer to the length of a road center line. Does this the refer to the potential network?
• On page 3, figure 1, the authors refer to the original OD points. I have to admit that the way information on user volumes is generated on basis of cell phone data, passed to point representation and finally is added as attribute of the line segment is not quite transparent the way it is currently described. (see also page 5)
• On page 4, reference is made to road section where steering selection is provided. What does this refer to?
• Also on page 4, node representation of public transport routes is described. While the authors point correctly to the transit stops of existing lines being relevant for consideration when selecting alternative routes and for network expansion, I am missing these stops here and in the following, when the routes constructed by the ants are evaluated. I would have expected that the evaluation should be better when stops can be reused.
• On page 5, section 3.1., I would have expected a more foundational reference rather than the ones given (22, 23).
• On page 6, the authors write that the ACO heuristics is intended to find an optimal solution (line 170, also figure 5). From my understanding, this is wrong. Indeed, it is capable of finding a feasible solution within finite times while the prove of solution being optimal cannot be made.
• On page 6, line 174: The meaning of the sentence is not quite clear to me.
• On page 6, before describing the approach used in detail, three introductional sentences about ACO would be helpful, especially with respect to the transitional rules and the role of the pheromone. Also, softmax method is introduced in figure 5 without explanation and reference. Further, the importance of incorporating a strategy for making the ants also explore prior unknown path alternatives has to be underlined somewhere here, as it prevents the heuristic to get stuck in local optima.
• On page 7, reference to non-linear coefficients is made several times – with it never be explained – is this your weighting factor? Also, abbreviations / notations used in the equations should be introduced earlier, e.g. equation 7 for 6.
• The typology-checking rule introduced on page 7 deserves a bit more elaboration given the consequence for accepted routing alternatives. On this page again, I would have expected to read something about the role of transit stops and their consideration when evaluation the constructed alternatives.
• On the beginning of Section 3.3, I would expect a reference for the relevance of the pheromone update procedure – there are a bunch of articles comparing different strategies, and the chosen one is neither motivated sufficiently nor cited.
• On page 9, results of a convergence analysis are presented, showing that roughly 400 iterations should be sufficient. I wonder why the authors then chose to use 1000 iterations for the results presented on the bottom of the page. On line 297, bus stops are mentioned again without their relevance becoming clear – neither is explained what travel analysis results is referred to here.
• On page 10, one solution constructed is presented. It would be interesting to see how much the constructed solutions over the course of the iterations are varying. On line 303ff, the authors attribute the positive results to the algorithm – I wonder to what extend it is actually due to the choice of the starting point…
• Table 1 is interesting but may profit from indicators referring to average occupancies or other user volume related aspects.
• On page 12, authors highlight that the proposed method is feasible for the generation of preliminary bus routes for new urban areas. From my understanding, this is specifically not the case (see results for area A), as user volumes, in this case being 0) are taking into account for the construction of alternatives.
• In the contribution section, the first 2 sentences are to be deleted.

Summing up, the article provides a good overview of the application of an ACO heuristics for public line planning but would profit from a minor revision prior to publishing.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Topic of the paper is interesting. Structure is well.

In my opinion literarute review is also enough.

I have one comment related to main goal of the proposed algorithm. Authors said "Along with the 109 goal of increasing the coverage of public transport lines, improving the operation effi-110 ciency of some routes is also considered". Please add inside the paper clearly criteria or goals which were implemented to algorithm. It will help reader to better understand the whole paper.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I'm satisfied with the revisions. No further comments.