Home

Welcome to the trashbots-RL Roadmap!

Reinforcement Learning Models to Try:

1. Encoding current bot as a coordinate instead of location in array.
2. Policy Gradient
3. Actor Critic Update

Environment Variants to Try:

1. Different distributions for trash i.e. coastal etc.
2. More than one item of trash per location
3. Impassable Terrain

Other Things to Try:

1. Return trash to deposit area

. . . .

continuous state space???

Current To-Do:

• Debugging (Evalfunctions: Action Counter / Memory, Trashsource closeness)
• Visualization
• Report