TRACKMANIARL / VISUAL GUIDE

The system, explained.

Nine diagrams covering training, demonstrations, value learning and game integration. Start with the runtime, then explore the part you need.

How off-policy training runs
GUIDE 01

How off-policy training runs

Open the diagram →
Resume the complete training state
GUIDE 02

Resume the complete training state

Open the diagram →
From demonstrations to a driving policy
GUIDE 03

From demonstrations to a driving policy

Open the diagram →
Compose a value-based agent
GUIDE 04

Compose a value-based agent

Open the diagram →
Align demonstrations with action time
GUIDE 05

Align demonstrations with action time

Open the diagram →
Learn from a complete replay history
GUIDE 06

Learn from a complete replay history

Open the diagram →
Understand the reward signal
GUIDE 07

Understand the reward signal

Open the diagram →
Choose a local or remote actor
GUIDE 08

Choose a local or remote actor

Open the diagram →
Connect the game to the agent
GUIDE 09

Connect the game to the agent

Open the diagram →