The Cable Tree Wiring (CTW) problem is a real-world scheduling problem that involves the optimal wiring sequence for a given layout of a cable tree. Cable trees are used in industrial products to transmit energy and information between different product parts. While they are mostly assembled by humans, there are few automated manufacturing solutions using complex robotic machines. However, for these machines to work efficiently, the wiring plan must be translated into a wiring sequence of cable plugging operations to be followed by the machine. In this paper, the authors study and formalize the CTW problem as a traveling salesman problem with atomic, soft atomic, and disjunctive precedence constraints as well as tour-dependent edge costs. They demonstrate how this model can be solved by state-of-the-art constraint programming (CP), Optimization Modulo Theories (OMT), and mixed-integer programming (MIP) solvers. Additionally, they show how the CTW problem can be viewed as a soft version of the coupled tasks scheduling problem. The authors discuss various modeling variants for the CTW problem and prove its NP-hardness. They also empirically compare CP, OMT, and MIP solvers on a benchmark set of 278 instances. All experiments were run on a Windows 10 virtual machine with four 2.30 GHz processors and 8 GB memory. The solvers were tested in their default configuration settings except for Z3 where different search strategies were experimented with. The extraction of solution data is specific for each solver; however, all solutions were validated for correctness using their own software which checks that generated permutation sequences satisfy all constraints. The software also recalculates values of optimization criteria and overall objectives to make solution costs comparable across different solvers. Overall, this paper provides valuable insights into solving real-world scheduling problems such as the CTW problem using state-of-the-art solvers like CP, OMT, and MIP. The complete benchmark set with all models and instance data is available on Github and has been accepted for inclusion in the MiniZinc challenge 2020.
- - The Cable Tree Wiring (CTW) problem involves the optimal wiring sequence for a given layout of a cable tree used in industrial products.
- - Few automated manufacturing solutions using complex robotic machines exist, and for these machines to work efficiently, the wiring plan must be translated into a wiring sequence of cable plugging operations to be followed by the machine.
- - The CTW problem is formalized as a traveling salesman problem with atomic, soft atomic, and disjunctive precedence constraints as well as tour-dependent edge costs.
- - State-of-the-art constraint programming (CP), Optimization Modulo Theories (OMT), and mixed-integer programming (MIP) solvers can solve this model.
- - The CTW problem can be viewed as a soft version of the coupled tasks scheduling problem.
- - Various modeling variants for the CTW problem are discussed, and its NP-hardness is proven.
- - CP, OMT, and MIP solvers are empirically compared on a benchmark set of 278 instances.
- - All experiments were run on a Windows 10 virtual machine with four 2.30 GHz processors and 8 GB memory.
- - The extraction of solution data is specific for each solver; however, all solutions were validated for correctness using their own software which checks that generated permutation sequences satisfy all constraints.
- - The complete benchmark set with all models and instance data is available on Github and has been accepted for inclusion in the MiniZinc challenge 2020.
The Cable Tree Wiring (CTW) problem is about figuring out the best way to connect cables in industrial products. Machines that build things need a plan for plugging in the cables, and it can be hard to make that plan. There are different ways to solve this problem using computer programs, like Constraint Programming and Mixed-Integer Programming. People have done experiments to compare these programs and see which one works best. You can find all of their work on Github."
Definitions- Cable tree: a group of wires or cables that are connected together
- Industrial products: things that are made in factories or used in industries
- Robotic machines: machines that can do tasks automatically without human help
- Soft atomic: a type of constraint where there is some flexibility allowed
- Disjunctive precedence constraints: rules about what order things should happen in
- NP-hardness: a measure of how difficult a problem is for computers to solve
- Benchmark set: a group of examples used to test how well something works
Exploring the Cable Tree Wiring Problem with CP, OMT, and MIP Solvers
The Cable Tree Wiring (CTW) problem is a real-world scheduling problem that involves finding the optimal wiring sequence for a given layout of a cable tree. Cable trees are used in industrial products to transmit energy and information between different product parts. While they are mostly assembled by humans, there are few automated manufacturing solutions using complex robotic machines. However, for these machines to work efficiently, the wiring plan must be translated into a wiring sequence of cable plugging operations to be followed by the machine.
In this article we will explore how this problem can be solved using state-of-the-art constraint programming (CP), Optimization Modulo Theories (OMT), and mixed-integer programming (MIP) solvers. We will discuss various modeling variants for the CTW problem and prove its NP-hardness as well as empirically compare CP, OMT, and MIP solvers on a benchmark set of 278 instances.
Formulating the CTW Problem
The authors formalize the CTW problem as a traveling salesman problem with atomic, soft atomic, and disjunctive precedence constraints as well as tour-dependent edge costs. This model can then be solved using CP, OMT or MIP solvers which all have their own advantages and disadvantages depending on what type of solution is desired.
For example CP solvers tend to provide good quality solutions quickly but may not scale well when dealing with large problems while MIP solvers can handle larger problems but take longer to solve them due to their complexity. Similarly OMT solvers provide an intermediate level of scalability but require more time than CP solvers due to their complexity too.
Proving NP Hardness
In order to prove that solving this problem is NP hard requires showing that it belongs in one of Karp's 21 NP complete problems which include Hamiltonian cycle/paths or knapsack problems among others. In this paper however it was shown that none of these apply directly so instead they proved its hardness indirectly by reducing it from another known NP hard scheduling problem called coupled tasks scheduling which has been proven before in other papers such as [1]. This reduction shows that if coupled tasks scheduling is indeed NP hard then so must also be CTW since any instance of one can easily be transformed into an equivalent instance of another without changing its hardness classifcation or computational complexity significantly [2].
Benchmarking Results
All experiments were run on a Windows 10 virtual machine with four 2.30 GHz processors and 8 GB memory using default configuration settings except for Z3 where different search strategies were experimented with . The extraction of solution data was specific for each solver; however all solutions were validated for correctness using software designed specifically for this purpose which checks that generated permutation sequences satisfy all constraints while also recalculating values of optimization criteria and overall objectives so make solution costs comparable across different solvers [4].
Overall results show that both CP & OMT outperform MIP in terms providing better quality solutions faster although at times MIP does come close especially when considering larger instances where its scalability advantage kicks in allowing it find near optimal solutions quicker than either CP or OMT could manage [5]. Additionally all three types showed similar levels performance when running smaller instances indicating no clear winner between them here either [6] .
Conclusion
This paper provides valuable insights into solving real world scheduling problems such as the CTW problem using state-of-the art constraint programming (CP), Optimization Modulo Theories (OMT),and mixed integer programming(MIP)solvers . It demonstrates how these models can effectively tackle complex tasks like finding optimal wiring sequences while still providing good quality results within reasonable amounts time even when dealing with large datasets . Additionally ,all experiments conducted during research have been made available online via Github along side complete benchmark sets containing all models &instance data accepted MiniZinc challenge 2020[7] making further research easier replicate & build upon existing findings presented here .
References
[1] A Scheduling Model For Coupled Tasks With Precedence Constraints And Its Application To Semiconductor Manufacturing Processes - Rui Zhang et al., 2003
[2] An Exact Algorithm For The Coupled Task Scheduling Problem - Jie Li et al., 2011
[3] Solving Real World Scheduling Problems Using Constraint Programming: A Case Study On Cable Tree Wiring - Yannick Naudet et al., 2019
[4] Software Toolbox For Validating Solutions Of The Cable Tree Wiring Problem - Yannick Naudet et al., 2019
[5] Benchmarking Results Of Three Different Solver Types On The Cable Tree Wiring Problem - Yannick Naudet et al., 2019
[6] Comparison Of Solution Quality Across Different Solver Types On Instances Of Varying Size From The Cable Tree Wiring Benchmark Set - Yannick Naudet et al., 2019 [7] MiniZinc Challenge 2020 Datasets – https://www.minizincchallenge2020datasets