From 02b20fbdb48f3064a54db41330e539b4dcf444d1 Mon Sep 17 00:00:00 2001 From: Johnny Fernandes Date: Sat, 25 Apr 2026 12:20:42 +0100 Subject: [PATCH] Sheep training flock _ improver --- training/herding_env.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/training/herding_env.py b/training/herding_env.py index 440d319..5fe5744 100644 --- a/training/herding_env.py +++ b/training/herding_env.py @@ -54,8 +54,9 @@ class HerdingEnv(gym.Env): # Reward weights (simple per-sheep progress — no phases, no gating) # ----------------------------------------------------------------------- W_PER_SHEEP = 2.0 # progress: sum of per-sheep distance-to-pen reductions - W_ALIGN = 0.05 # dog on anti-pen side of COM — directional hint only, - # kept tiny so sit-still is never profitable vs completion + W_ALIGN = 0.0 # disabled: created a sit-still trap from n_sheep≥2. + # Progress reward already encodes "be on anti-pen side" + # implicitly (sheep flee toward pen → positive progress). W_PEN_BONUS = 10.0 # per sheep penned W_COMPLETE = 100.0 # all sheep penned W_STEP_COST = 0.02 # time penalty — strong enough to punish doing nothing