A small hybrid B&B + CP‑SAT demo using a slack‑based probe controller Relative Feasibility

[Devonlangford55]

A minimal FJSP demo showing how a slack‑based controller, Relative Feasibility (RF), decides when to run CP‑SAT inside a Branch‑and‑Bound search. RF acts as a safety‑verified decision layer: at each node it computes a slack term (gap between the node’s lower bound and the incumbent) and a time_factor (how much time remains relative to expected work). Their product,

rf_score = slack × time_factor,

determines whether the node should be probed using a fast greedy completion or an optional CP‑SAT interval model. The implementation is conservative, monotonic‑timed, and uses CP‑SAT strictly as a verifier. The demo includes greedy preseeding, SCAT machine ordering, strict‑mode pruning, and full trace outputs (JSON + CSV)
rf_fjsp_ortools_demo_Version4.py