WAYFINDERRL / 01
✳ REINFORCEMENT LEARNING OBSERVATORY
Intelligence, in motion.
A world to explore. A policy to discover. Watch an agent learn its own way.
Archipelago
21 × 17 TERRAIN · 210 WALKABLE CELLSSelect an object, then place it on the terrain. Use Move to reposition existing objects.
Episodes completed 0/ 2,000
Mean reward —R
Last 50 episodes
Beacon reached —%
Success · last 50 episodes
Exploration 100.0%
0 environment steps
The learning curve
Episode reward20-episode mean
A learning curve begins with a little exploration.Start training to watch the reward evolve.
Inside the policy
CELL 2, 14↑North0.00
↑East0.00
↑South0.00
↑West0.00
LaunchSelect an object to place it at the coordinates below.
Inspect cell
THE LEARNING LOOP
One decision at a time.
Inspect the latest real transition.
01 / OBSERVEThe worldCurrent state
02 / ACTChoose a moveExplore or exploit
03 / RECEIVEA rewardEnvironment feedback
04 / UPDATERefine a valueTemporal difference error