Classical and Quantum Heuristics for the Binary Paint Shop Problem

arXiv:2509.15294 · quant-ph, cs.DS, cs.ET, math.OC · Submitted 2025-09-18 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Classical and Quantum Heuristics for the Binary Paint Shop Problem".

Mira: The gist The level-1 XQAOA ansatz with parameters optimised by gradient-based methods achieves a limiting paint-swap ratio of approximately 0.357 Problem and Formulation The Binary Paint Shop Problem (BPSP) is…

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So, we started by introducing this Binary Paint Shop Problem as a hard optimization challenge in car manufacturing where you're trying to paint two cars of each model with different colors while keeping those color swaps between consecutive cars low >

Mira: The authors set out to prove that they can tackle this by reducing it to a weighted MaxCut problem and then further down to an Ising spin model, which is what lets them use quantum algorithms like QAOA >

Kai: Their main claim is that they formalize this reduction using the Initial Car Colour encoding, which cuts down on the number of decision variables we have to manage in the problem >

Mira: They then use this structure to derive exact interaction strengths from the BPSP constraints and prove some new structural properties about these resulting graphs, including a maximum degree bound >

Lev: From my viewpoint, proving those structural properties is crucial because it tells us exactly what kind of graph we're dealing with before we even try to run any quantum circuit on it >

Kai: The paper then moves on to compare two state-of-the-art low-depth QAOA variants, XQAOA and RQAOA, against several classical heuristics like Red-First and Recursive Greedy >

Mira: They find that XQAOA1 achieves an average paint swap ratio of zero point three five seven, which is better than RQAOA1 and all four classical baselines they tested across instances from two cars up to twelve cars >

Lev: That performance gap suggests that for the small-depth circuits they investigated, the expressiveness in XQAOA provides a clear advantage over the recursive approach of RQAOA >

Kai: The core message here is that XQAOA1 is stable across all tested sizes, meaning it doesn't suddenly degrade as the problem size increases, unlike RQAOA1 which falls behind RSG at larger instances >

Mira: It matters because it suggests that if we encode these kinds of optimization problems into Ising models, using a well-parameterized low-depth ansatz might give you a reliable performance benchmark >

Lev: For anyone looking at this for actual quantum hardware implementation, the stability of XQAOA1 is a big deal because it implies you don't need to constantly retune your parameters as the problem size grows >

Conclusion: Kai: So, to wrap up "Classical and Quantum Heuristics for the Binary Paint Shop Problem", we're looking at how classical methods compare against these new quantum heuristics like XQAOA1 >

Mira: The authors essentially show that by reducing the paint swap problem to MaxCut, they can use quantum tools to get a result, and that XQAOA1 is consistently performing better than what we see from traditional recursive heuristics across various sizes >

Lev: What this means for us in the real world is that if you have a complex scheduling or resource allocation problem that fits this structure, starting with an expressive QAOA ansatz seems like a more reliable path than just using a greedy approach >

Kai: It highlights that for problems mapped to Ising models, the structure of your quantum circuit, its expressiveness, seems to be a bigger factor than how deep you make it when you're keeping the depth low >

Mira: And the paper also lays out a very concrete mathematical way to calculate what the optimal cost is for these BPSP instances using that Ising formulation >

Lev: That final formula gives us a clear target—it tells us exactly what we are trying to minimize, which is useful whether you're running it on a simulator or trying to build it into an actual quantum device >

Centre for Quantum Technologies (CQT), National University of Singapore · Quantum Innovation Centre (Q.InC), Agency for Science, Technology and Research (A*STAR) · Centre for Quantum Computation and Communication Technologies (CQC2T), Department of Quantum Science and Technology, Australian National University · Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR) · Science, Mathematics and Technology Cluster, Singapore University of Technology and Design

quant-ph, cs.DS, cs.ET, math.OC

Submitted: 2025-09-18

Updated: 2026-10-09

Comments: 32 pages, 3 figures. Final revised version accepted for publication in Physical Review Applied

DOI: 10.1103/dw96-xxfy

Code: https://github.com/vijeycreative/XQAOA-BPSP

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: The gist The level-1 XQAOA ansatz with parameters optimised by gradient-based methods achieves a limiting paint-swap ratio of approximately 0.357 Problem and Formulation The Binary Paint Shop Problem

Key concepts

Binary Paint Shop Problem (BPSP): This is an optimization task in car manufacturing where you must assign one of two colors to each car. The constraint is that for every model, the two cars of that model must have different colors. The goal is to minimize the total number of times a color changes between adjacent cars.
Paint Swap Ratio: This metric measures production efficiency by counting the average number of color changes per car. A lower ratio means fewer swaps and higher efficiency, which is what the algorithms aim to minimize.
Quantum Approximate Optimisation Algorithm (QAOA): QAOA is a quantum algorithm used to find near-optimal solutions for combinatorial problems like BPSP. The paper tests two versions of this algorithm: XQAOA1 and RQAOA1, to see which performs better on real-world instances.

Terminology

Binary Paint Shop Problem — 2× in this paper
Benchmark — 1× in this paper · explained in 2 episodes
Quantum Approximate Optimisation Algorithm — 1× in this paper
Recursive Star Greedy — 1× in this paper

Summary

The gist The level-1 XQAOA ansatz with parameters optimised by gradient-based methods achieves a limiting paint-swap ratio of approximately 0.357

Problem and Formulation

The Binary Paint Shop Problem (BPSP) is an APX-hard optimisation problem in automotive manufacturing where the task is to decide which of two colours to paint each car so that the two occurrences of each model are painted differently, while minimising consecutive colour swaps. The key performance metric is the paint swap ratio, defined as the average number of colour changes per car. The paper formalizes this problem through a reduction to weighted MaxCut and subsequently to an Ising spin model using an Initial Car Colour (ICC) encoding that halves the number of decision variables. This rigorous foundation allows for the application of QAOA and its variants to BPSP.

Classical Heuristics vs. Quantum Algorithms

The study benchmarks two state-of-the-art low-depth QAOA variants, eXpressive QAOA (XQAOA) and Recursive QAOA (RQAOA), against classical heuristics such as Red-First, Greedy, Recursive Greedy (RG), and Recursive Star Greedy (RSG). The numerical results show that XQAOA1 achieves an average swap ratio of 0.357, surpassing RQAOA1 and all known classical heuristics, including the conjectured performance of RSG. In contrast, RQAOA1 shows diminishing performance as the problem size increases, falling behind RSG for 211-car instances and on all 212-car instances. XQAOA1 stays robust across all sizes, indicating strong potential to asymptotically surpass all known heuristics.

Performance Trends at Scale

The benchmarks reveal a clear divergence in performance as the problem size increases. While Red-First scales the worst, Recursive-Star Greedy is the most effective classical heuristic with an average ratio of about 0.37, above its conjectured value of 0.361. XQAOA1 attains an average ratio of roughly 0.357, outperforming all tested classical heuristics. RQAOA1 exhibits size-dependent performance: it is competitive on smaller instances, but its average ratio rises steadily with n, eventually falling behind both XQAOA1 and Recursive-Star Greedy. The mean of the RQAOA1 distributions drifts steadily rightward from 0.322 at n = 128 to 0.390 at n = 4096, with the right tail extending to 0.4, matching the performance of Recursive Greedy.

Theoretical Connections and Reductions

The BPSP problem is exactly equivalent to solving weighted MaxCut on a graph derived from the BPSP instance. The total weight of the BPSP graph Gx can be expressed compactly in terms of the double letter count and the eta function. Furthermore, a polynomial-time reduction from BPSP to weighted MaxCut is established, showing that minimizing paint swaps is exactly equivalent to solving weighted MaxCut on the BPSP graph. This equivalence leads to Corollary 15, which states that BPSP(x) = ξn (σRF(x)) − MaxCut(Gx).

Conclusion and Conjecture

The findings establish XQAOA1 as the most robust method tested at scale, with its distributions sharply peaked around 0.357. This consistency motivates Conjecture 1, which states that for Binary Paint Shop Problem instances encoded as Ising models, the level-1 XQAOA ansatz with parameters optimised by gradient-based methods achieves a limiting paint-swap ratio of approximately 0.357. The results indicate that expressiveness may be a more influential factor than non-locality in shaping the performance of QAOA-type heuristics for general Ising models. The gap between the conjectured asymptotic performance of XQAOA1 and the lower bound on the optimal solution suggests that further algorithmic improvements remain possible. The paper concludes by establishing a polynomial-time reduction from BPSP to weighted MaxCut, enabling efficient solving of BPSP instances using a MaxCut oracle. The final result is expressed in Corollary 18, which provides formulas for the optimal cost and set of optimal solutions based on the Ising Hamiltonian. A key insight is that BPSP(x) = n − 1/2 + 1/2 Wtot (Gx) − MaxCut (Gx). The final formulation shows that BPSP∗(x) = MaxCut∗(Gx). This provides a polynomial-time reduction from BPSP to weighted MaxCut problem, as demonstrated in Algorithm 2. A key result is that the optimal cost and the set of optimal solutions of the BPSP problem are, respectively, 1/2 + x − 1 − max z∈[-1,1]xX−1 i=1 (−1)η(x,i)zxi zxi+1 and 1/2 + x − 1 + dl(x) − max z∈[-1,1]xX−1 i=1 xi̸=xi+1 (−2)η(x,i)zxi zxi+1. The final expression for the optimal cost in the Ising formulation is H(z1,..., zn) = − 2/Xn-1 i=1 xi̸=xi+1 (−1)η(x,i)zxi zxi+1. This entire framework provides a rigorous foundation for applying QAOA and its variants to BPSP. The paper is supported by the Agency for Science, Technology and Research (A star) under the Central Research Fund (CRF) Award for Use-Inspired Basic Research (UIBR), A star C230917003, and the Quantum Innovation Centre (Q.InC) Strategic Research and Translational Thrust. The code used to generate the numerical results, together with the datasets (including the random BPSP instances used in our benchmarks), are publicly available at github.com/vijeycreative/XQAOA-BPSP. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust. The paper is supported by A star C230917003 and Q.InC Strategic Research and Translational Thrust.

Improvements for AI systems

Improve quantum optimization for combinatorial problems by utilizing level-1 XQAOA to achieve a stable limiting performance of approximately 0.357 for Binary Paint Shop Problem instances encoded as Ising models, as conjectured in Conjecture 1: "For Binary Paint Shop Problem instances encoded as Ising models, the level-1 XQAOA ansatz with parameters optimised by gradient-based methods achieves a limiting paint-swap ratio of approximately 0.357, i.e. limn→∞ E[∆C/n] ≈ 0.357."
Develop a quantum heuristic for BPSP that leverages the structure of the problem via reduction to weighted MaxCut, specifically by solving weighted MaxCut on the BPSP graph to find MaxCut(Gx) and z∗ ∈ MaxCut∗(Gx). This allows an AI system to compute the optimal solution by using Algorithm 2: Polynomial-time reduction from BPSP (Problem 1) to MaxCut (Problem 2).
Implement a robust method for finding the optimal paint swap ratio by employing the relationship derived in Corollary 15, which states that BPSP(x) = ξn (σRF(x)) − MaxCut(Gx) = n − 1/2 + 1/2 ldl(x) + 1/2 Wtot (Gx) − MaxCut (Gx). This provides a polynomial-time reduction to solve the BPSP problem by querying the weighted MaxCut oracle for Problem 2 on Gx.
Create an efficient classical heuristic for BPSP that utilizes the structural properties of the BPSP graph, specifically by finding a cut that maximizes wtGx(z), as this is equivalent to solving MaxCut(G) = max z ∈ r, b n wtG(z) on the corresponding graph G.
Design an iterative algorithm for refining quantum parameters in variational algorithms that accounts for non-local dependencies, by adopting the approach of Recursive QAOA (RQAOA1) which uses QAOA1 as its subroutine to recursively reduce the problem size until it becomes tractable for brute-force methods.