search

opendilab / DI-engine

OpenDILab Decision AI Engine. The Most Comprehensive Reinforcement Learning Framework B.P.
https://di-engine-docs.readthedocs.io
Apache License 2.0
2.97k stars 364 forks source link

Agent Demo List #3

Open PaParaZz1 opened 3 years ago

PaParaZz1 commented 3 years ago

This issue is a collection of various interesting agent demonstration trained by DI-engine, it will be updated continually.

zjowowen commented 1 year ago

https://user-images.githubusercontent.com/93968541/207054782-6cea7a97-9a0d-41bb-8d58-4484c877c6e0.mp4

zjowowen commented 1 year ago

Evogym Carrier-v0 random bot VS trained agent

https://user-images.githubusercontent.com/93968541/210569397-ad55fd85-71ee-48a3-889d-93e3e85ebf4f.mp4

agent during early training VS trained agent

https://user-images.githubusercontent.com/93968541/210812515-bb4a10af-b319-4238-89bf-514a0a4a473c.mp4

zjowowen commented 1 year ago

Procgen Bigfish random bot VS trained agent

https://user-images.githubusercontent.com/93968541/212252949-250e96fd-cc03-4565-aae0-355c85ffe7f1.mp4

timothijoe commented 1 year ago

Metadrive env with PPO algorithm demonstration. Map Sequence: Straight + Intersection + Straight + Roundabout

https://user-images.githubusercontent.com/49814804/218450981-4a72e55d-ec4d-4980-b52b-3bef6dbc72a0.mp4

zjowowen commented 9 months ago

cliffwalking agent trained by DQN random agent vs trained agent

https://github.com/opendilab/DI-engine/assets/93968541/756cbba6-4405-4882-8753-6e4288e40a40

phuongboi commented 8 months ago

Hi @zjowowen, thank you for your sharing. I would like to ask about fly through gate simulation, did you change the compute reward and terminal function or you use the same with original work.