INITIALIZING NEURAL AGENT…
🤖 AI ROBOT LEARNING SIMULATOR
Q-Learning · Three.js / WebGL
Controls
Learning Speed
5
Environment Difficulty
Easy (6x6, few obstacles)
Medium (9x9, moderate)
Hard (12x12, dense)
Insane (14x14, moving hazards)
Robot Move Speed
1.0x
Sensor Range
2
Exploration (ε start)
1.0
⏸ Pause
↺ Reset Learning
⟲ New Environment
🎥 Follow Cam
〰 Toggle Path
✦ Toggle Trail
Legend:
Green pillar = goal · Red blocks = obstacles · Cyan trail = visited path · Yellow line = current optimal policy path.
Drag to orbit · Scroll to zoom · Right-drag to pan.
Performance
Episode
0
Step (this episode)
0
Total Steps
0
Episode Reward
0
Avg Reward (20 ep)
0
Best Episode Reward
-∞
Successes
0
Success Rate
0%
Collisions
0
Learning Rate α
0.30
Exploration ε
1.00
Neural Decision Map
Inputs: obstacle sensors (N/E/S/W) + goal vector → hidden layer → action confidence (softmax of Q-values). Brightest output edge = chosen action.
Reward History
Per-episode total reward (grey) and 20-episode moving average (cyan). Trend rising = agent is learning.