Pusher-v5 PPO 학습 시각화 리포트
Gymnasium MuJoCo Robotic Arm Manipulation • Stable-Baselines3 PPO
Total Steps: 301,056
Total Timesteps
301,056
Best Return
-30.84
Final Return
-30.84
Best Obj-to-Goal Dist
0.069 m
학습 단계별 행동 변화 (Visual Progressions)
Baseline (Step 0)
Step: 0
Mean Reward
-57.51
Obj-Goal Dist
0.220 m
Checkpoint (30,000 steps)
Step: 30,000
Mean Reward
-49.77
Obj-Goal Dist
0.220 m
Checkpoint (60,000 steps)
Step: 60,000
Mean Reward
-43.23
Obj-Goal Dist
0.190 m
Checkpoint (90,000 steps)
Step: 90,000
Mean Reward
-38.93
Obj-Goal Dist
0.191 m
Checkpoint (120,000 steps)
Step: 120,000
Mean Reward
-42.36
Obj-Goal Dist
0.222 m
Checkpoint (150,000 steps)
Step: 150,000
Mean Reward
-41.61
Obj-Goal Dist
0.237 m
Checkpoint (180,000 steps)
Step: 180,000
Mean Reward
-36.00
Obj-Goal Dist
0.145 m
Checkpoint (210,000 steps)
Step: 210,000
Mean Reward
-34.57
Obj-Goal Dist
0.142 m
Checkpoint (240,000 steps)
Step: 240,000
Mean Reward
-32.33
Obj-Goal Dist
0.107 m
Checkpoint (270,000 steps)
Step: 270,000
Mean Reward
-32.52
Obj-Goal Dist
0.115 m
Checkpoint (300,000 steps)
Step: 300,000
Mean Reward
-31.46
Obj-Goal Dist
0.084 m
Final (301,056 steps)
Step: 301,056
Mean Reward
-30.84
Obj-Goal Dist
0.069 m
학습 수렴 및 메트릭 분석 차트
