mlsanityData to 14 Oct 2026

Geometry-Aware Reinforcement Learning for 2D Irregular Nesting

arXiv 2606.10611 ↗First posted 9 Jun 2026cs.LG cs.CVcited by 0

On the map: Deep reinforcement learning methods → RL with diffusion and flow models

Abstract (arXiv)

Traditional heuristic solvers for the 2D irregular nesting problem share a fundamental limitation: they are blind to polygon geometry, relying on guided brute-force to navigate the continuous placement space with minimal geometrical guidance. In this paper, we argue that Reinforcement Learning is uniquely positioned to overcome this bottleneck. By pairing an optimization policy with a geometry-aware neural encoder, an agent can automatically discover rich geometric priors directly from data, utilizing these learned intuitions to strategically guide exploration. To realize this, we introduce the Polygons Transformer (PoT), a novel architecture that encodes 2D continuous vector geometries while allowing cross-polygon attention. We couple this novel architecture with a Combinatorial Optimization Reinforcement Learning (CORL) training framework to find optimal solutions. To support this paradigm, we release an open-source training dataset derived from complex geographic contours alongside a dedicated evaluation benchmark. Empirically, our agent slightly exceeds Sparrow, the state-of-the-art heuristic, on small (4-polygon) instances, while a clear scaling gap remains on larger (8-polygon) instances.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in RL with diffusion and flow models, one of 15 directions in Deep reinforcement learning methods. Builds on 20 of its 30 references that are in the map. First posted on arXiv 9 Jun 2026.

Builds on

20 of 30 references.

References · 20

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2017Proximal Policy Optimization Algorithmscited by 15,215
  4. 2020Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domainscited by 2,200
  5. 2015Pointer Networkscited by 895
  6. 2018Attention, Learn to Solve Routing Problems!cited by 656
  7. 2016Neural Combinatorial Optimization with Reinforcement Learningcited by 608
  8. 2020POMO: Policy Optimization with Multiple Optima for Reinforcement Learningcited by 348
  9. 2021Learning to Iteratively Solve Routing Problems with Dual-Aspect Collaborative Transformercited by 108
  10. 2023RL4CO: an Extensive Reinforcement Learning for Combinatorial Optimization Benchmarkcited by 71
  11. 2025Robust Autonomy Emerges from Self-Playcited by 67
  12. 2022Winner Takes It All: Training Performant RL Populations for Combinatorial Optimizationcited by 58
8 more
  1. 2023Combinatorial Optimization with Policy Adaptation using Latent Space Searchcited by 41
  2. 2024PolyNet: Learning Diverse Solution Strategies for Neural Combinatorial Optimizationcited by 35
  3. 2022Towards General-Purpose Representation Learning of Polygonal Geometriescited by 22
  4. 2024PolygonGNN: Representation Learning for Polygonal Geometries with Heterogeneous Visibility Graphcited by 10
  5. 2023Learning based 2D Irregular Shape Packingcited by 5
  6. 2024GFPack++: Improving 2D Irregular Packing by Learning Gradient Field with Attentioncited by 2
  7. 2025An open-source heuristic to reboot 2D nesting researchcited by 2
  8. 2025Decoupling Geometry from Optimization in 2D Irregular Cutting and Packing Problems: an Open-Source Collision Detection Enginecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Playing to Par: Reinforcement Learning for Provably Optimal Quadrilateral Block Decompositions
  2. 2026TriSearch: Learning to Optimize Triangulations via Bistellar Flips
  3. 2026Flow3D-OPD: Multi-Teacher On-Policy Distillation for 3D Geometry Generation with Flow-Matching Diffusion Transformer
  4. 2026Dmsh: A Multi-Agent Reinforcement Learning Framework for All-Quad Mesh Generation
  5. 2026Control-Geometry Straightening for Sampling-Based Latent Planning
  6. 2026GeoRouteNet: A Geometry-Aware Non-Autoregressive Neural Solver for the Euclidean Traveling Salesman Problem
  7. 2025RLCAD: Reinforcement Learning Training Gym for Revolution Involved CAD Command Sequence Generation
  8. 2026GeoPAR: Large-Scale Multi-Agent Combinatorial Optimization with Geometry-Guided Parallel Autoregressive Learning