GridWorld Playground

Select a GridWorld from the menu below, or draw one yourself!

• Use '*' for walls (all rows must be the same width)
• Use 'g' to indicate the goal state(s)
• Use 's' to indicate a start state (or leave unspecified to always start at a different state)

4 rooms Maze Puddle world Square 4x4 (with start state) Square 10x10 Start learning!
************* * * g* * * * * * * * * * * * ** **** * * *** *** * * * * * * * * * * * *************

These agents are learning using standard Q-learning:
In this playground you can:

• Change all the learning/environment parameters in real-time
• Click anywhere in the top GridWorld to position the agent there
• The cell colours will change to reflect the agent's current value estimate
• Transfer values between the top/bottom agents