Greedy Approaches for Packing While Travelling with Deterministic and Stochastic Constraints
Pith reviewed 2026-05-10 12:13 UTC · model grok-4.3
The pith
New reward functions for greedy packing outperform standard methods in the packing while travelling problem under both fixed and uncertain item weights.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The authors establish that reward functions tailored to the packing while travelling problem, when used in greedy algorithms and extended through a hyper-heuristic framework, deliver better solutions than standard heuristics; the same functions adapted for stochastic weights via chance constraints retain this advantage.
What carries the argument
Tailored reward functions that score items by combining profit, weight, and tour-dependent travel cost, embedded in a hyper-heuristic selector and adapted to chance-constrained stochastic weights.
If this is right
- Improved packing solutions directly raise the quality of complete travelling thief problem solvers that alternate between tour construction and packing.
- The hyper-heuristic layer reduces the need for manual tuning of reward functions when moving between different instance classes.
- Chance-constraint handling allows the same greedy framework to be deployed in settings where item weights fluctuate, such as variable cargo densities.
- Runtime remains comparable to standard greedy methods, supporting repeated calls inside larger metaheuristics.
Where Pith is reading between the lines
- The approach could be transferred to other problems that interleave routing and resource selection, such as vehicle routing with loading constraints.
- If the reward functions prove robust, they might serve as drop-in replacements in existing travelling thief solvers without requiring changes to the tour component.
- Further work could test whether the same functions remain effective when the tour itself is allowed to change slightly during packing.
Load-bearing premise
The new reward functions and hyper-heuristic choices deliver consistent gains across the full range of problem sizes, item distributions, and tour structures that arise in practice.
What would settle it
A large-scale test set of packing while travelling instances where standard greedy heuristics achieve equal or higher profit than the proposed tailored functions on at least half the instances would refute the claimed benefit.
Figures
read the original abstract
The travelling thief problem (TTP) is a well-known multi-component optimisation problem that captures the interdependence between two components: the tour across cities and the packing of items. The packing while travelling problem (PWT) is an NP-hard subproblem of TTP where the packing of items should be optimised for a given fixed tour. In many solvers, the packing component is often addressed using greedy heuristics. Here, the use of suitable greedy functions is essential for the success of greedy algorithms. In this paper, we introduce new reward functions tailored to the PWT and extend them to a hyper-heuristic framework to achieve further advantage. Furthermore, we investigate the chance constrained PWT for greedy approaches and adopt the newly introduced reward functions for stochastic weights. The experimental results clearly demonstrate the benefit of the tailored heuristics over the standard heuristics in both deterministic and stochastic constraints.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces new reward functions tailored to the Packing While Travelling (PWT) subproblem of the Travelling Thief Problem, extends them via a hyper-heuristic framework, and adapts the approach to chance-constrained stochastic weights. It claims that experimental comparisons demonstrate clear superiority of these tailored greedy heuristics over standard ones (e.g., profit/weight) under both deterministic and stochastic settings.
Significance. If the experimental claims hold under rigorous verification, the work could offer practical improvements to TTP solvers by strengthening the packing component with problem-specific greedy rules and hyper-heuristics, especially for stochastic variants where standard methods may underperform.
major comments (3)
- [Experimental Results] The experimental results section provides no details on the instance sets (city counts, item distributions, tour qualities), statistical tests, or derivation of the new reward functions, preventing verification that the reported advantages are robust rather than instance-specific (as flagged in the abstract's superiority claim).
- [Experimental Results] Runtime comparisons do not appear to normalize or account for the hyper-heuristic framework's selection overhead against the baseline greedy methods, which could undermine the claim that the framework adds value without hidden computational costs.
- [Stochastic PWT] For the stochastic (chance-constrained) case, it is unclear how probability estimation is integrated into the greedy choice without introducing bias or excessive computation, and whether this is consistently applied across all compared heuristics.
minor comments (2)
- [Methods] Notation for the new reward functions could be clarified with explicit equations early in the methods section to aid reproducibility.
- [Introduction] The abstract and introduction would benefit from a brief statement on the specific standard heuristics used as baselines for direct comparison.
Simulated Author's Rebuttal
We thank the referee for the thorough review and constructive feedback. We address each of the major comments below and will make revisions to improve the clarity and completeness of the manuscript.
read point-by-point responses
-
Referee: [Experimental Results] The experimental results section provides no details on the instance sets (city counts, item distributions, tour qualities), statistical tests, or derivation of the new reward functions, preventing verification that the reported advantages are robust rather than instance-specific (as flagged in the abstract's superiority claim).
Authors: We agree that more detailed information on the experimental setup is necessary for full verification and reproducibility. In the revised manuscript, we will expand Section 4 to include: a table summarizing the instance characteristics (number of cities from 20 to 280 as per standard TTP instances, item distributions with varying knapsack capacities), details on tour generation methods (using Lin-Kernighan heuristic for TSP component), and the full derivation of the reward functions in Section 3 with mathematical motivations. Additionally, we will report results of statistical significance tests (e.g., paired t-tests or Wilcoxon tests with p-values) across multiple runs. This will demonstrate that the advantages are not instance-specific. revision: yes
-
Referee: [Experimental Results] Runtime comparisons do not appear to normalize or account for the hyper-heuristic framework's selection overhead against the baseline greedy methods, which could undermine the claim that the framework adds value without hidden computational costs.
Authors: The hyper-heuristic framework's selection process is lightweight, involving a simple rule-based or learning-based selector that evaluates a small set of heuristics per decision point. However, we acknowledge the need for transparent runtime analysis. In the revision, we will provide normalized runtime results, including separate timings for the greedy decisions and the hyper-heuristic overhead. We will show that the overhead is negligible compared to the overall solving time and that the improved solution quality justifies it. This will be added to the experimental results section. revision: yes
-
Referee: [Stochastic PWT] For the stochastic (chance-constrained) case, it is unclear how probability estimation is integrated into the greedy choice without introducing bias or excessive computation, and whether this is consistently applied across all compared heuristics.
Authors: For the chance-constrained stochastic PWT, the probability estimation is performed using a fixed number of Monte Carlo simulations (1000 samples per evaluation) to approximate the probability that the total weight does not exceed the capacity under stochastic weights. This estimation is incorporated into the reward function by multiplying the original reward by the estimated probability of feasibility, thus guiding the greedy choice towards items that maintain high reward while satisfying the constraint with high probability. This approach is applied uniformly to all heuristics, including the standard profit/weight and the new tailored ones, to ensure fair comparison. We will add a new subsection in the stochastic section detailing the integration, including pseudocode and analysis of computational overhead, which remains acceptable as the sampling is efficient. revision: yes
Circularity Check
No circularity: empirical heuristic comparison with independent experimental validation
full rationale
The paper is an empirical study that introduces tailored reward functions for greedy heuristics on the Packing While Travelling (PWT) problem and evaluates them experimentally against standard baselines under deterministic and chance-constrained stochastic settings. No derivation chain exists; the central claims rest on runtime and solution-quality comparisons across problem instances rather than any self-referential equations, fitted parameters renamed as predictions, or load-bearing self-citations. The reward functions are defined directly from PWT structure (tour-dependent profit/time trade-offs) and then tested, with no reduction of outputs to inputs by construction. This is a standard, self-contained empirical contribution.
Axiom & Free-Parameter Ledger
Reference graph
Works this paper leans on
-
[1]
Approximate approaches to the traveling thief problem
Hayden Faulkner, Sergey Polyakovskiy, Tom Schultz, and Markus Wagner. Approximate approaches to the traveling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 385–392. ACM, 2015
work page 2015
-
[2]
Joseph B Kruskal. On the shortest spanning subtree of a graph and the traveling salesman problem.Proceedings of the American Mathematical society, 7(1):48–50, 1956
work page 1956
-
[3]
A note on two problems in connexion with graphs.Numerische mathematik, 1(1):269–271, 1959
Edsger W Dijkstra. A note on two problems in connexion with graphs.Numerische mathematik, 1(1):269–271, 1959
work page 1959
-
[4]
Randomized greedy algorithms for covering problems
Wanru Gao, Tobias Friedrich, Frank Neumann, and Christian Hercher. Randomized greedy algorithms for covering problems. InProceedings of the Genetic and Evolutionary Computation Conference, page 309–315. ACM, 2018
work page 2018
-
[5]
Submodular function maximization
Andreas Krause and Daniel Golovin. Submodular function maximization. InTractability, pages 71–104. Cambridge University Press, 2014. 11 APREPRINT- APRIL16, 2026
work page 2014
-
[6]
Mohammad Reza Bonyadi, Zbigniew Michalewicz, and Luigi Barone. The travelling thief problem: The first step in the transition from theoretical problems to realistic problems. In2013 IEEE Congress on Evolutionary Computation, pages 1037–1044. IEEE, 2013
work page 2013
-
[7]
The packing while traveling problem.Eur
Sergey Polyakovskiy and Frank Neumann. The packing while traveling problem.Eur. J. Oper. Res., 258(2):424– 439, 2017
work page 2017
-
[8]
Analysis of baseline evolutionary algorithms for the packing while travelling problem
Vahid Roostapour, Mojgan Pourhassan, and Frank Neumann. Analysis of baseline evolutionary algorithms for the packing while travelling problem. InProceedings of the 15th ACM/SIGEVO Conference on Foundations of Genetic Algorithms, page 124–132. ACM, 2019
work page 2019
-
[9]
Socially inspired algorithms for the traveling thief problem
Mohammad Reza Bonyadi, Zbigniew Michalewicz, Michał Roman Przybyłek, and Adam Wierzbicki. Socially inspired algorithms for the traveling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 421–428. ACM, 2014
work page 2014
-
[10]
Exact approaches for the travelling thief problem
Junhua Wu, Markus Wagner, Sergey Polyakovskiy, and Frank Neumann. Exact approaches for the travelling thief problem. InProceedings of the Simulated Evolution and Learning - 11th International Conference, volume 10593 ofLecture Notes in Computer Science, pages 110–121. Springer, 2017
work page 2017
-
[11]
Mohamed El Yafrani and Belaïd Ahiod. Efficiently solving the traveling thief problem using hill climbing and simulated annealing.Information Sciences, 432:231–244, 2018
work page 2018
-
[12]
Investigation of the traveling thief problem
Rogier Hans Wuijts and Dirk Thierens. Investigation of the traveling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 329–337. ACM, 2019
work page 2019
-
[13]
Evolutionary multitasking for the scenario-based travelling thief problem
Thilina Pathirage Don, Aneta Neumann, and Frank Neumann. Evolutionary multitasking for the scenario-based travelling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, page 809–817. ACM, 2025
work page 2025
-
[14]
Mohamed El Yafrani and Belaïd Ahiod. Population-based vs. single-solution heuristics for the travelling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 317–324. ACM, 2016
work page 2016
-
[15]
Majid Namazi, M. A. Hakim Newton, Abdul Sattar, and Conrad Sanderson. A profit guided coordination heuristic for travelling thief problems. InProceedings of the 12th International Symposium on Combinatorial Search, pages 140–144. AAAI Press, 2019
work page 2019
-
[16]
Efficient hybrid local search heuristics for solving the travelling thief problem
Alenrex Maity and Swagatam Das. Efficient hybrid local search heuristics for solving the travelling thief problem. Appl. Soft Comput., 93:106284, 2020
work page 2020
-
[17]
Zitong Zhang, Lei Yang, Peipei Kang, Xiaotian Jia, and Wensheng Zhang. Solving the traveling thief problem based on item selection weight and reverse-order allocation.IEEE Access, 9:54056–54066, 2021
work page 2021
-
[18]
Marcella S. R. Martins, Mohamed El Yafrani, Myriam R. B. S. Delgado, Markus Wagner, Belaïd Ahiod, and Ricardo Lüders. Hseda: a heuristic selection approach based on estimation of distribution algorithm for the travelling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, page 361–368. ACM, 2017
work page 2017
-
[19]
Mohamed El Yafrani, Marcella Martins, Markus Wagner, Belaïd Ahiod, Myriam Delgado, and Ricardo Lüders. A hyperheuristic approach based on low-level heuristics for the travelling thief problem.Genetic Programming and Evolvable Machines, 19(1–2):121–150, 2018
work page 2018
-
[20]
Hyper-heuristic approaches for the travelling thief problem
Fathelrahman Ali and Mohamedelfatih Mohamedkhair. Hyper-heuristic approaches for the travelling thief problem. InInternational Conference on Computer, Control, Electrical, and Electronics Engineering, pages 1–6. IEEE, 2021
work page 2021
-
[21]
A sequence-based hyper- heuristic for traveling thieves.Applied Sciences, 13(1):56, 2022
Daniel Rodríguez, Jorge M Cruz-Duarte, José Carlos Ortiz-Bayliss, and Ivan Amaya. A sequence-based hyper- heuristic for traveling thieves.Applied Sciences, 13(1):56, 2022
work page 2022
-
[22]
Tamalika Sarkar and Chandrasekharan Rajendran. Travelling thief problem: a survey of recent variants, so- lution approaches and future directions.International Journal of Systems Science: Operations & Logistics, 11(1):2424200, 2024
work page 2024
-
[23]
Chance constrained programming with joint constraints.Operations Research, 13:930–945, 1965
Bruce L Miller and Harvey M Wagner. Chance constrained programming with joint constraints.Operations Research, 13:930–945, 1965
work page 1965
-
[24]
Evolutionary algorithms for the chance-constrained knapsack problem
Yue Xie, Oscar Harper, Hirad Assimi, Aneta Neumann, and Frank Neumann. Evolutionary algorithms for the chance-constrained knapsack problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 338–346. ACM, 2019
work page 2019
-
[25]
Aneta Neumann, Yue Xie, and Frank Neumann. Evolutionary algorithms for limiting the effect of uncertainty for the knapsack problem with stochastic profits. InParallel Problem Solving from Nature - 17th International 12 APREPRINT- APRIL16, 2026 Conference, Proceedings, Part I, volume 13398 ofLecture Notes in Computer Science, pages 294–307. Springer, 2022
work page 2026
-
[26]
Chance-constrained multiple- choice knapsack problem: Model, algorithms, and applications.IEEE Trans
Xuanfeng Li, Shengcai Liu, Jin Wang, Xiao Chen, Yew-Soon Ong, and Ke Tang. Chance-constrained multiple- choice knapsack problem: Model, algorithms, and applications.IEEE Trans. Cybern., 54(12):7969–7980, 2024
work page 2024
-
[27]
Benjamin Doerr, Carola Doerr, Aneta Neumann, Frank Neumann, and Andrew M. Sutton. Optimization of chance-constrained submodular functions. InProceedings of the 34th AAAI Conference on Artificial Intelligence, pages 1460–1467. AAAI Press, 2020
work page 2020
-
[28]
Optimizing chance-constrained submodular problems with variable uncertainties
Xiankun Yan, Anh Viet Do, Feng Shi, Xiaoyu Qin, and Frank Neumann. Optimizing chance-constrained submodular problems with variable uncertainties. InProceedings of the 26th European Conference on Artificial Intelligence, volume 372 ofFrontiers in Artificial Intelligence and Applications, pages 2826–2833. IOS Press, 2023
work page 2023
-
[29]
Aneta Neumann and Frank Neumann. Optimizing monotone chance-constrained submodular functions using evolutionary multiobjective algorithms.Evolutionary Computation, 33(3):363–393, 2025
work page 2025
-
[30]
Aneta Neumann, Jakob Bossek, and Frank Neumann. Diversifying greedy sampling and evolutionary diversity optimisation for constrained monotone submodular functions. InProceedings of the Genetic and Evolutionary Computation Conference, pages 261–269. ACM, 2021
work page 2021
-
[31]
Sampling-based pareto optimization for chance-constrained monotone submodular problems
Xiankun Yan, Aneta Neumann, and Frank Neumann. Sampling-based pareto optimization for chance-constrained monotone submodular problems. InProceedings of the Genetic and Evolutionary Computation Conference, pages 621–629. ACM, 2024
work page 2024
-
[32]
In-time agent-based vehicle routing with a stochastic improvement heuristic
Robert Kohout, Kutluhan Erol, and C Robert. In-time agent-based vehicle routing with a stochastic improvement heuristic. InProceedings of the 16th National Conference on Artificial Intelligence, pages 864–869. AAAI, 1999
work page 1999
-
[33]
Dynamic vehicle routing with stochastic requests
Russell Bent and Pascal Van Hentenryck. Dynamic vehicle routing with stochastic requests. InProceedings of the 18th International Joint Conference on Artificial Intelligence, pages 1362–1363. Morgan Kaufmann, 2003
work page 2003
-
[34]
Waiting and relocation strategies in online stochastic vehicle routing
Russell Bent and Pascal Van Hentenryck. Waiting and relocation strategies in online stochastic vehicle routing. InProceedings of the 20th International Joint Conference on Artificial Intelligence, pages 1816–1821. Morgan Kaufmann, 2007
work page 2007
-
[35]
Feng Shi, Xiankun Yan, and Frank Neumann. Runtime analysis of simple evolutionary algorithms for the chance-constrained makespan scheduling problem. InParallel Problem Solving from Nature - 17th International Conference, Proceedings, Part II, volume 13399 ofLecture Notes in Computer Science, pages 526–541. Springer, 2022
work page 2022
-
[36]
The chance constrained travelling thief problem: Problem formulations and algorithms
Thilina Pathirage Don, Aneta Neumann, and Frank Neumann. The chance constrained travelling thief problem: Problem formulations and algorithms. InProceedings of the Genetic and Evolutionary Computation Conference, pages 214–222. ACM, 2024
work page 2024
-
[37]
Weighted-scenario optimisation for the chance constrained travelling thief problem
Thilina Pathirage Don, Aneta Neumann, and Frank Neumann. Weighted-scenario optimisation for the chance constrained travelling thief problem. InIEEE Congress on Evolutionary Computation, pages 1–8. IEEE, 2025
work page 2025
-
[38]
Marín-Blázquez, Sonia Schulenburg, and Emma Hart
Peter Ross, Javier G. Marín-Blázquez, Sonia Schulenburg, and Emma Hart. Learning a procedure that can solve hard bin-packing problems:A new ga-based approach to hyper-heuristics. InGenetic and Evolutionary Computation Conference. Proceedings, Part II, volume 2724 ofLecture Notes in Computer Science, pages 1295–1306. Springer, 2003
work page 2003
-
[39]
A comprehensive benchmark set and heuristics for the traveling thief problem
Sergey Polyakovskiy, Mohammad Reza Bonyadi, Markus Wagner, Zbigniew Michalewicz, and Frank Neumann. A comprehensive benchmark set and heuristics for the traveling thief problem. InProceedings of the Genetic and Evolutionary Computation Conference, pages 477–484. ACM, 2014
work page 2014
-
[40]
Chained lin-kernighan for large traveling salesman problems
David Applegate, Wiliam Cook, and Andre Rohe. Chained lin-kernighan for large traveling salesman problems. INFORMS Journal on Computing; Winter, 15:82–92, 2003. 13
work page 2003
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.