DriveAI — Training

DriveAI

›

Training Lab

Phase 2

Mission Challenge Learn Home

Reward Events

Training Track

Evaluation track is always fixed.

Reward History

Metrics

Press Train to start learning. Metrics update in real-time.

Episode—

Current Reward—

Best Reward—

Avg Reward—

Distance—

Collisions—

Q-Updates—

States Explored—

Exploration ε—

Best Episodes

No episodes yet.

Current Discrete State

—

Action Values

Run training or simulation to see Q-values.

About Q-Values

Each number shows how much total reward the agent expects from choosing that action in this state. Higher = the agent has learned that action works well here.

Policy History

Train to see how the policy changes over time.

Watch a saved episode

Before vs After Training

Use Replay First Episode to see untrained behaviour, then Replay Best Episode after training to compare improvement.

Speed: Slow Fast Medium

Connect and click Train to begin.