Research map / Reinforcement learning / Visual and representation RL
Online 3D Bin Packing with Constrained Deep Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use constrained DRL with prediction-and-projection rather than the listed online heuristics | Our BPP-1 · CUT-1 | space utilization: 51.9% -> 73.4% | Helped |
| Use constrained DRL with prediction-and-projection rather than the listed online heuristics | Our BPP-1 · CUT-2 | space utilization: 49.2% -> 66.9% | Helped |
| Use constrained DRL with prediction-and-projection rather than LBP | Our BPP-1 · RS | space utilization: 54.7% -> 50.5% | Hurt |
| Add mask prediction, mask constraint, and feasibility-based entropy together (ablation) | BPP-1 · CUT-2 | space utilization: 7.82% -> 66.9% | Helped |
| Add mask constraint using projection (ablation) | BPP-1 with MP and FE · CUT-2 | space utilization: 27.9% -> 66.9% | Helped |
| Add feasibility-based entropy (ablation) | BPP-1 with MP and MC · CUT-2 | space utilization: 63.7% -> 66.9% | Helped |
| Add mask prediction (ablation) | BPP-1 with MC and FE · CUT-2 | space utilization: 63.0% -> 66.9% | Helped |
| Allow item re-orientation (ablation) | BPP-1 · RS | space utilization: 50.5% -> 62.1% | Helped |
| Allow item re-orientation (ablation) | BPP-1 · CUT-1 | space utilization: 73.4% -> 76.2% | Helped |
| Allow item re-orientation (ablation) | BPP-1 · CUT-2 | space utilization: 66.9% -> 70.2% | Helped |
| Replace step-wise reward with termination reward (ablation) | BPP-1 · RS | space utilization: 50.5% -> 39.2% | Hurt |
| Replace step-wise reward with termination reward (ablation) | BPP-1 · CUT-1 and CUT-2 | space utilization: 73.4% -> 72.5%; 66.9% -> 66.5% | No clear effect |
Closest papers by meaning
- Adjustable Robust Reinforcement Learning for Online 3D Bin Packing (NeurIPS 2023 · similarity 0.92)
- Online Learning in CMDPs: Handling Stochastic and Adversarial Constraints (ICML 2024 · similarity 0.83)
- Near-Optimal Sample Complexity for Online Constrained MDPs (NeurIPS 2025 · similarity 0.82)
- Distributionally Robust Multi-Agent Reinforcement Learning for Dynamic Chute Mapping (ICML 2025 · similarity 0.82)
- Structured Reinforcement Learning for Combinatorial Decision-Making (NeurIPS 2025 · similarity 0.82)
- Learning Multimodal Behaviors from Scratch with Diffusion Policy Gradient (NeurIPS 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.