Imitation learning lifecycle and controlled RL handoff Behavior cloning is the supervised algorithm; imitation_learning owns the wider data, recovery, evaluation and warm-start workflow. Trusted demonstration data Behavior-cloning lifecycle Closed-loop gate and RLexperiment training set held-out laps epochs candidate metrics pass fail / intervene record-demo whole laps ·outlier filtering DAgger recovery student states ·expert labels Dataset quality complete laps ·timing Dataset manifest fingerprint ·provenance Lap-level split train / validation;no leakage Safe augmentation horizontal flip;preserve unknowns BehaviorCloningModel encoder · temporal ·actor head Offline learner contiguous batches ·AMP · scaler Validationaggregate loss numerator /sample count BC checkpoint exact trainer state Exact resume same split, orderand next update Closed-loop eval finish · pace ·intervention rate Promotion gate named criteria, neverimplicit Warm-start RL encoder + temporal;new value head Package boundary: trackmaniarl.trackmania.imitation_learning. Public classes include LidarBehaviorCloningModelFactory andBehaviorCloningLearner. Warm-start is partial loading with an explicit match report; it is not a resume and it does not transfer optimizer state or episodichidden state.