Updating for new flagrun again

Author: sash-a

configs/flagrun.json

@@ -5,14 +5,15 @@
     "ant_env_rew_weight": 1,
     "path_rew_weight": 0,
     "dist_rew_weight": 0,
-    "goal_reach_rew": 5000,
+    "goal_reach_rew": 500,
     "kwargs": {
       "enclosed": true,
       "timeout": 100,
       "size": 10,
       "max_target_dist": 4,
       "max_targets": 0,
-      "tolerance": 0.5
+      "tolerance": 1,
+      "switch_flag_on_collision": false
     }
   },
   "noise": {

flagrun.py

@@ -125,9 +125,10 @@ def run_model(model: PrimFF,
                 if render:
                     env.render('human')
                     # time.sleep(1 / 100)
-                    env.stadium_scene._p.addUserDebugLine([*pos, 0.5], [*(pos + ob[:2]), 0.5], lifeTime=0.1)
+                    # env.stadium_scene._p.addUserDebugLine([*pos, 0.5], [*(pos + ob[:2]), 0.5], lifeTime=0.1)
                     # robot to goal
-                    # env.stadium_scene._p.addUserDebugLine(pos, [env.walk_target_x, env.walk_target_y, pos[2]], lifeTime=0.1)
+                    env.stadium_scene._p.addUserDebugLine([*pos, 0.5], [env.walk_target_x, env.walk_target_y, 0.5],
+                                                          lifeTime=0.1)
                     # robot dir
                     # point = [10, m * 10 + c, pos[2]]
                     # env.stadium_scene._p.addUserDebugLine([x, y, pos[2]], point, lifeTime=0.1, lineColorRGB=[0, 1, 0])
@@ -178,6 +179,7 @@ def r_fn(model: PrimFF, use_ac_noise=True) -> TrainingResult:
                                            rs if use_ac_noise else None, False)
         return RewardResult(rews, behv, obs if save_obs else np.array([np.zeros(env.observation_space.shape)]), steps)
 
+
     assert cfg.general.policies_per_gen % comm.size == 0 and (cfg.general.policies_per_gen / comm.size) % 2 == 0
     eps_per_proc = int((cfg.general.policies_per_gen / comm.size) / 2)
     for gen in range(cfg.general.gens):  # main loop