pith. sign in

arxiv: 2604.08296 · v1 · submitted 2026-04-09 · 💻 cs.NE

Robust Multi-Objective Optimization for Bicycle Rebalancing in Shared Mobility Systems

Diego Daniel Pedroza-Perez , Gabriel Luque , Sergio Nesmachnow , Jamal Toutouh This is my paper

Pith reviewed 2026-05-10 17:29 UTC · model grok-4.3

classification 💻 cs.NE
keywords bike-sharingrebalancingmulti-objective optimizationNSGA-IIdemand uncertaintyrobustnessshared mobilitygenetic algorithms
0
0 comments X

The pith

A tri-objective genetic algorithm generates robust overnight rebalancing routes for bike-sharing that balance travel distance, average unmet demand, and performance under high-demand spikes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper frames static overnight truck rebalancing in dock-based bike-sharing as a tri-objective problem that simultaneously minimizes total travel distance, expected unmet demand, and a robustness measure focused on high-demand scenarios. Solutions are evolved with NSGA-II using a permutation-partition encoding and custom relocation operators, then evaluated by recourse simulation that enforces truck loads and station capacities across many demand realizations. Experiments on the real Barcelona Bicing network of 460 stations produce well-distributed Pareto sets that contribute substantially to the reference non-dominated front, while greedy constructive baselines mostly produce extreme points that are frequently dominated.

Core claim

The central claim is that Non-dominated Sorting Genetic Algorithm II, equipped with a permutation-partition representation and domain-specific operators including a biased best-improvement relocation move, approximates high-quality trade-off sets for the tri-objective static rebalancing problem under uncertainty, as validated by substantial coverage of the reference Pareto set in experiments on real Barcelona data with 460 stations.

What carries the argument

The tri-objective optimization model with recourse simulation for robustness evaluation, solved via NSGA-II using permutation-partition encoding and biased best-improvement relocation operators.

Load-bearing premise

The demand uncertainty model and recourse simulation accurately capture real high-demand variability and enforce truck loads and station capacities across realizations without introducing bias in the robustness measure.

What would settle it

Re-running the full NSGA-II procedure and greedy baselines on a different bike-sharing network or on updated Barcelona demand data and checking whether the evolutionary Pareto sets remain better distributed and contribute more to the reference front than the baselines.

Figures

Figures reproduced from arXiv: 2604.08296 by Diego Daniel Pedroza-Perez, Gabriel Luque, Jamal Toutouh, Sergio Nesmachnow.

Figure 1
Figure 1. Figure 1: Example of solution encoding of a scenario with 9 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: BB1-MAX operator diagram. The red arrows indi [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Instance map: the red points are the bike-shared [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: 2D front projection: distance vs. critical unmet de [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 6
Figure 6. Figure 6: Three-dimensional Pareto front obtained by the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Two-dimensional Reference Pareto front projection: [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 11
Figure 11. Figure 11: Unmet demand distribution of the solutions of our [PITH_FULL_IMAGE:figures/full_fig_p008_11.png] view at source ↗
Figure 10
Figure 10. Figure 10: Total distance distribution of the solutions of our [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗
Figure 12
Figure 12. Figure 12: Heatmap of pairwise comparisons based on the [PITH_FULL_IMAGE:figures/full_fig_p012_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Heatmap of pairwise comparisons based on the [PITH_FULL_IMAGE:figures/full_fig_p012_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Heatmap of pairwise comparisons based on the [PITH_FULL_IMAGE:figures/full_fig_p012_14.png] view at source ↗
read the original abstract

Dock-based bike-sharing systems exhibit spatial imbalances between bicycle supply and user demand, often addressed through overnight truck-based rebalancing. This work studies static overnight rebalancing under demand uncertainty modeled as a tri-objective optimization problem. The objectives minimize total travel distance, expected unmet demand, and a robustness-oriented unmet demand measure over high-demand scenarios. Route plans are evaluated via a recourse simulation that enforces truck loads and station capacity constraints across multiple demand realizations. The robustness objective supports selecting plans that reduce peak-demand service degradation. Trade-off solutions are approximated with Non-dominated Sorting Genetic Algorithm II using a permutation--partition encoding and domain-specific relocation operators, including a biased best-improvement move for station relocation. Experiments on the real Barcelona Bicing system with 460 stations show well-distributed Pareto sets and substantial contributions to the reference non-dominated set. Greedy constructive baselines mainly yield extreme solutions and are often dominated.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper formulates static overnight rebalancing in dock-based bike-sharing systems as a tri-objective optimization problem under demand uncertainty. Objectives minimize total travel distance, expected unmet demand, and a robustness-oriented unmet demand measure over high-demand scenarios. Plans are optimized via NSGA-II using a permutation-partition encoding and domain-specific relocation operators, then evaluated through a recourse simulation enforcing truck loads and station capacities across realizations. Experiments on the real Barcelona Bicing system (460 stations) report well-distributed Pareto sets with substantial contributions to the reference non-dominated set, while greedy baselines mainly produce extreme and often dominated solutions.

Significance. If the demand uncertainty model holds, the work provides a practical evolutionary multi-objective framework for generating rebalancing plans that explicitly trade off operational cost, average service level, and robustness to peak-demand degradation in shared mobility systems. The application to a large real-world instance with simulation-based evaluation and baseline comparisons offers concrete evidence of the value of the tri-objective formulation over single-objective or greedy approaches.

major comments (2)
  1. [Method / Demand Modeling] Demand uncertainty model and high-demand scenario generation (described in the method section and abstract): no quantitative validation is provided that the generated high-demand scenarios reproduce the tail behavior, spatial correlations, or observed variability of real Barcelona peak-hour demand. This directly affects the robustness objective and thus the reported Pareto distribution and dominance claims versus greedy baselines, as the simulation may favor certain route structures without empirical grounding.
  2. [Evaluation / Recourse Simulation] Recourse simulation details (evaluation procedure): the abstract and method description provide no explicit quantitative checks on how unmet demand is aggregated across realizations or whether the enforcement of truck loads and capacities introduces systematic bias into the robustness metric. This is load-bearing for the central empirical claim of well-distributed Pareto sets on the 460-station instance.
minor comments (2)
  1. [Abstract] The abstract could state the number of demand realizations used and the precise mathematical definition of the robustness objective to improve clarity.
  2. [NSGA-II Operators] Clarify the implementation of the biased best-improvement move for station relocation and the exact permutation-partition encoding to support reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review of our manuscript. We address each major comment below and describe the revisions we will make to improve the clarity and empirical grounding of the work.

read point-by-point responses

  1. Referee: [Method / Demand Modeling] Demand uncertainty model and high-demand scenario generation (described in the method section and abstract): no quantitative validation is provided that the generated high-demand scenarios reproduce the tail behavior, spatial correlations, or observed variability of real Barcelona peak-hour demand. This directly affects the robustness objective and thus the reported Pareto distribution and dominance claims versus greedy baselines, as the simulation may favor certain route structures without empirical grounding.

    Authors: We agree that quantitative validation of the high-demand scenarios is important for supporting the robustness objective and the resulting Pareto analysis. The manuscript describes scenario generation from historical Barcelona Bicing demand data but does not include explicit statistical comparisons. In the revised manuscript we will add a new subsection (or appendix) reporting quantitative checks, including comparisons of variance, spatial correlation matrices, and tail quantiles between the generated scenarios and observed peak-hour realizations. These additions will directly address the potential impact on the reported trade-off solutions and baseline comparisons. revision: yes

  2. Referee: [Evaluation / Recourse Simulation] Recourse simulation details (evaluation procedure): the abstract and method description provide no explicit quantitative checks on how unmet demand is aggregated across realizations or whether the enforcement of truck loads and capacities introduces systematic bias into the robustness metric. This is load-bearing for the central empirical claim of well-distributed Pareto sets on the 460-station instance.

    Authors: We concur that additional detail on the recourse simulation procedure and aggregation of unmet demand is needed to ensure transparency and to rule out systematic bias. The current description states that unmet demand is evaluated across realizations while enforcing truck capacities and station limits, but does not provide quantitative diagnostics. In the revision we will expand the evaluation section to specify the aggregation function (average unmet demand across realizations), include a sensitivity study on the number of realizations used, and discuss how capacity constraints are enforced to avoid biasing the robustness metric. These clarifications will strengthen the evidence for the well-distributed Pareto sets. revision: yes

Circularity Check

0 steps flagged

No significant circularity; formulation and results are self-contained via external data and standard MOEA.

full rationale

The paper defines a tri-objective problem (minimize travel distance, expected unmet demand, and robustness over high-demand scenarios) and solves it with NSGA-II using a permutation-partition encoding and domain-specific operators. Route plans are evaluated through an independent recourse simulation enforcing truck loads and capacities across demand realizations drawn from the Barcelona Bicing system. No equation or claim reduces by construction to a fitted parameter, self-citation chain, or renamed input; the Pareto sets and dominance claims versus greedy baselines arise directly from the optimization on external realizations rather than internal redefinition. The approach is therefore non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard assumptions from stochastic optimization and evolutionary algorithms; no free parameters or invented entities are explicitly introduced beyond the problem formulation.

axioms (1)
  • domain assumption Demand uncertainty can be adequately represented by a finite set of realizations for recourse simulation.
    The evaluation of objectives depends on simulating multiple demand scenarios to compute expected and robust unmet demand.

pith-pipeline@v0.9.0 · 5459 in / 1268 out tokens · 40277 ms · 2026-05-10T17:29:28.573800+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    Parallel metaheuris- tics: recent advances and new trends.International Transactions in Operational Research, 20(1):1–48, 2013

    Enrique Alba, Gabriel Luque, and Sergio Nesmachnow. Parallel metaheuris- tics: recent advances and new trends.International Transactions in Operational Research, 20(1):1–48, 2013

    work page 2013

  2. [2]

    Performance indicators in multiobjective optimization.European Journal of Operational Research, 292(2):397–422, 2021

    Charles Audet, Jean Bigeon, Dominique Cartier, Sébastien Le Digabel, and Lu- dovic Salomon. Performance indicators in multiobjective optimization.European Journal of Operational Research, 292(2):397–422, 2021

    work page 2021

  3. [3]

    Blank and K

    J. Blank and K. Deb. pymoo: Multi-objective optimization in python.IEEE Access, 8:89497–89509, 2020

    work page 2020

  4. [4]

    OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks, September 2017

    Geoff Boeing. OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks, September 2017

    work page 2017

  5. [5]

    A survey and analysis of evolutionary operators for permuta- tions

    Vincent Cicirello. A survey and analysis of evolutionary operators for permuta- tions. InProceedings of the 15th International Joint Conference on Computational Intelligence - ECTA, pages 288–299. INSTICC, SciTePress, 2023

    work page 2023

  6. [6]

    Multiobjective electric vehicle charg- ing station locations in a city scale area: Malaga study case

    Christian Cintrano and Jamal Toutouh. Multiobjective electric vehicle charg- ing station locations in a city scale area: Malaga study case. In Juan Luis Jiménez Laredo, J. Ignacio Hidalgo, and Kehinde Oluwatoyin Babaagba, editors, Applications of Evolutionary Computation, pages 584–600, Cham, 2022. Springer International Publishing

    work page 2022

  7. [7]

    Deb.Multi-Objective Optimization using Evolutionary Algorithms

    K. Deb.Multi-Objective Optimization using Evolutionary Algorithms. John Wiley & Sons, 2001

    work page 2001

  8. [8]

    The bike sharing rebalancing problem with stochastic demands.Transportation Research Part B: Methodological, 118:362–380, 2018

    Mauro Dell’Amico, Manuel Iori, Stefano Novellani, and Anand Subramanian. The bike sharing rebalancing problem with stochastic demands.Transportation Research Part B: Methodological, 118:362–380, 2018

    work page 2018

  9. [9]

    Hongfei Jia, Hongzhi Miao, Guangdong Tian, MengChu Zhou, Yixiong Feng, Zhiwu Li, and Jiangchen Li. Multiobjective bike repositioning in bike-sharing systems via a modified artificial bee colony algorithm.IEEE Transactions on Automation Science and Engineering, 17(2):909–920, 2020

    work page 2020

  10. [10]

    Spread assessment for evolutionary multi-objective optimization

    Miqing Li and Jinhua Zheng. Spread assessment for evolutionary multi-objective optimization. InEvolutionary Multi-Criterion Optimization: 5th International Conference, EMO 2009, Nantes, France, April 7-10, 2009. Proceedings 5, pages 216–

    work page 2009

  11. [11]

    The static bike rebalancing problem with optimal user incentives.Transportation Research Part E: Logistics and Transportation Review, 146:102216, 2021

    Yanfeng Li and Yang Liu. The static bike rebalancing problem with optimal user incentives.Transportation Research Part E: Logistics and Transportation Review, 146:102216, 2021

    work page 2021

  12. [12]

    The irace package: Iterated racing for automatic algorithm configuration.Operations Research Perspectives, 3:43–58, 2016

    Manuel López-Ibáñez, Jérémie Dubois-Lacoste, Leslie Pérez Cáceresand Mauro Bi- rattari, and Thomas Stützle. The irace package: Iterated racing for automatic algorithm configuration.Operations Research Perspectives, 3:43–58, 2016

    work page 2016

  13. [13]

    Comparison of multiobjective evolutionary algorithms for prioritized urban waste collection in Montevideo, Uruguay.Electronic Notes in Discrete Mathematics, 69:93–100, 2018

    Sergio Nesmachnow, Diego Gabriel Rossit, and Jamal Toutouh. Comparison of multiobjective evolutionary algorithms for prioritized urban waste collection in Montevideo, Uruguay.Electronic Notes in Discrete Mathematics, 69:93–100, 2018

    work page 2018

  14. [14]

    Neumann-Saavedra and Rossana Cavagnini

    Bruno A. Neumann-Saavedra and Rossana Cavagnini. Column-generation-based heuristics for integrating static rebalancing and faulty bike collection in bike- sharing systems.OR Spectrum, 47(2):375–409, Jun 2025

    work page 2025

  15. [15]

    Aliki Pouliasi, Amalia Nikolopoulou, Konstantinos Gkiotsalitis, and Konstantinos Kepaptsoglou. An exact approach for the static docked bike rebalancing problem that minimizes the routing costs and unmet demand under various geographic considerations.Journal of Public Transportation, 27:100141, 2025

    work page 2025

  16. [16]

    Tal Raviv, Michal Tzur, and Iris A. Forma. Static repositioning in a bike-sharing system: models and solution approaches.EURO Journal on Transportation and Logistics, 3(2):175–203, 2013

    work page 2013

  17. [17]

    Shui and W.Y

    C.S. Shui and W.Y. Szeto. A review of bicycle-sharing service planning problems. Transportation Research Part C: Emerging Technologies, 117:102648, 2020

    work page 2020

  18. [18]

    Soft computing methods for multiobjective location of garbage accumulation points in smart cities.Annals of Mathematics and Artificial Intelligence, 88(1):105–131, 2020

    Jamal Toutouh, Diego Rossit, and Sergio Nesmachnow. Soft computing methods for multiobjective location of garbage accumulation points in smart cities.Annals of Mathematics and Artificial Intelligence, 88(1):105–131, 2020

    work page 2020

  19. [19]

    Vallez, Mario Castro, and David Contreras

    Carlos M. Vallez, Mario Castro, and David Contreras. Challenges and opportuni- ties in dock-based bike-sharing rebalancing: A systematic review.Sustainability, 13(4), 2021

    work page 2021

  20. [20]

    Multiobjective optimization using evolutionary algorithms—a comparative case study

    Eckart Zitzler and Lothar Thiele. Multiobjective optimization using evolutionary algorithms—a comparative case study. InInternational conference on parallel problem solving from nature, pages 292–301. Springer, 1998. GECCO ’26, July 13-17, 2026, San José, Costa Rica Diego Daniel Pedroza-Perez, Gabriel Luque, Sergio Nesmachnow, and Jamal Toutouh A Greedy c...

    work page 1998