Reinforcement Learning

Learning placement/sequencing policies from reward signals.

Also called: Reinforcement Learning · RL · 심층 강화학습

A learning-based approach in which a policy for selecting and placing pieces is learned from a reward signal (e.g. utilization). Packing and nesting results have been reported with RL-based methods under specific benchmark conditions; this is not a claim of state-of-the-art or production readiness. See the evidence policy for how such unverified claims are phrased.

Claims & evidence

Every relationship is a claim with an evidence grade; applicable problem-to-problem relationships also carry an equivalence level. See the evidence policy.

A–X describes evidence reliability; E0–E4 describes mathematical similarity between problems. The two axes are independent.

Relationships to this entry 2

uses method
2D Irregular Nesting to Reinforcement Learning
RL-based nesting has been reported in recent literature, but its industrial maturity and generalization are not verified here. [unverified — no grade-A source recorded]
EvidenceD · Expert inference
Sources
No source listed
uses method
3D Bin Packing to Reinforcement Learning
RL-based 3D packing/loading has been reported recently, but its industrial maturity is not verified here. [unverified]
EvidenceD · Expert inference
Sources
No source listed

Relationships to this entry
Relationship	Claim	Evidence	Equiv.	Sources
uses method 2D Irregular Nesting to Reinforcement Learning	RL-based nesting has been reported in recent literature, but its industrial maturity and generalization are not verified here. [unverified — no grade-A source recorded]	D · Expert inference	—	No source listed
uses method 3D Bin Packing to Reinforcement Learning	RL-based 3D packing/loading has been reported recently, but its industrial maturity is not verified here. [unverified]	D · Expert inference	—	No source listed