Skip to main content
QUICK REVIEW

[论文解读] Domain-Independent Dynamic Programming with Constraint Propagation

Imko Marijnissen, J. Christopher Beck|arXiv (Cornell University)|Mar 17, 2026
Constraint Satisfaction and Optimization被引用 0
一句话总结

该论文将约束编程中的约束传播整合到 Domain-Independent Dynamic Programming (DIDP) 中,以剪裁状态和转移,从而在三个组合问题上提高解的覆盖率。

ABSTRACT

There are two prevalent model-based paradigms for combinatorial problems: 1) state-based representations, such as heuristic search, dynamic programming (DP), and decision diagrams, and 2) constraint and domain-based representations, such as constraint programming (CP), (mixed-)integer programming, and Boolean satisfiability. In this paper, we bridge the gap between the DP and CP paradigms by integrating constraint propagation into DP, enabling a DP solver to prune states and transitions using constraint propagation. To this end, we implement constraint propagation using a general-purpose CP solver in the Domain-Independent Dynamic Programming framework and evaluate using heuristic search on three combinatorial optimisation problems: Single Machine Scheduling with Time Windows, the Resource Constrained Project Scheduling Problem (RCPSP), and the Travelling Salesperson Problem with Time Windows (TSPTW). Our evaluation shows that constraint propagation significantly reduces the number of state expansions, causing our approach to solve more instances than a DP solver for Single Machine Scheduling and RCPSP, and showing similar improvements for tightly constrained TSPTW instances. The runtime performance indicates that the benefits of propagation outweigh the overhead for constrained instances, but that further work into reducing propagation overhead could improve performance further. Our work is a key step in understanding the value of constraint propagation in DP solvers, providing a model-based approach to integrating DP and CP.

研究动机与目标

  • 通过在一个基于模型的 DP 框架中嵌入约束传播来弥合 DP 与 CP 的差距。
  • 提供使用 DP(基于状态)与 CP(基于约束)表示的对偶视角,以实现剪裁和更强的对偶界。
  • 开发一个通用的、基于模型的集成,适用于任何形式的约束传播,同时支持启发式搜索。

提出的方法

  • 通过一个通用的 CP 求解器在 DIDP 中实现约束传播。
  • 使用带有 CP 导出的传播步骤的对偶界信息的 DP 框架,产生 D' 并收紧对偶界。
  • 用 GenSucc(S) 替换为 GenSuccPropagation(S, primal) 并更新 Dual(S) 以纳入 DualCP(S, D'),在状态与后继层面对不可行性进行检查。
  • 建立三种 DP 模型 (1|r_i, δ_i|∑ w_i T_i, RCPSP, TSPTW) 并展示 CP 传播器 (Disjunctive, Cumulative) 如何剪裁搜索空间。
  • 在 RPID 中评估 A* 与 Complete Anytime Beam Search (CABS),有无约束传播的对比,以及与基于 CP 的 OR-Tools 求解器的对比。
Figure 1: An overview of the interactions in our framework.
Figure 1: An overview of the interactions in our framework.

实验结果

研究问题

  • RQ1将约束传播整合到 DIDP 是否会比单独的 DP 求解器在测试问题上减少状态扩展数量?
  • RQ2CP 驱动的传播是否在更严格的实例约束下提高了能解决更多实例的能力,传播开销对运行时间有何影响?
  • RQ3传播如何影响状态和后继层面的对偶界与不可行性检测?
  • RQ4在三种基准问题(带时间窗的单机调度、RCPSP、TSPTW)中,哪一种在不同约束紧度下最能从传播中受益?
  • RQ5将通用的 CP 传播与基于模型的 DP 框架结合时,会有哪些实际性能权衡?

主要发现

  • 约束传播在三大问题上显著减少了状态扩展数量。
  • 在 1|r_i, δ_i|∑ w_i T_i 和 RCPSP 上,该方法解决的实例数量多于仅 DP 求解器,在约束更紧的 TSPTW 实例上也有改进。
  • 对于高度受约束的实例,运行时间有所提升,但传播开销表明需要进一步降低传播时间以获得额外收益。
  • CABS+CP 在充足运行时间后通常优于 CABS 的已解决实例数量,尽管初始传播存在开销。
  • 与 OR-Tools CP-SAT 相比,DP+传播的组合显示出有效性,凸显了基于模型的 DP-CP 集成的价值。
(a) $1|r_{i},\delta_{i}|\sum w_{i}T_{i}$ instances solved compared to state expansions. Our approach is able to solve the most instances per expansion.
(a) $1|r_{i},\delta_{i}|\sum w_{i}T_{i}$ instances solved compared to state expansions. Our approach is able to solve the most instances per expansion.

更好的研究,从现在开始

从论文设计到论文写作,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。