Overcooked Demo

A web application where humans can play Overcooked with trained AI agents and replay trajectories.

• Installation
• Usage
• Dependencies
• Using Pre-trained Agents
• Updating Overcooked_ai
• Configuration
• Legacy Code

Installation

Building the server image requires Docker

Usage

The server can be deployed locally using the driver script included in the repo. To run the production server, use the command

./up.sh --env production

In order to build and run the development server, which includes a deterministic scheduler and helpful debugging logs, run

./up.sh

After running one of the above commands, navigate to http://localhost

In order to kill the production server, run

./down.sh

Dependencies

The Overcooked-Demo server relies on both the overcooked-ai and human-aware-rl repos. The former contains the game logic, the latter contains the rl training code required for managing agents. Both repos are automatically cloned and installed in the Docker builds.

The branch of overcooked_ai and human_aware_rl imported in both the development and production servers can be specified by the --overcooked-branch and --harl-branch parameters, respectively. For example, to use the branch foo from overcooked-ai and branch bar from human_aware_rl, run

./up.sh --overcooked-branch foo --harl-branch bar

The default branch for both repos is currently master.

Using Pre-trained Agents

Overcooked-Demo can dynamically load pre-trained agents provided by the user. In order to use a pre-trained agent, a pickle file should be added to the agents directory. The final structure will look like static/assets/agents/<agent_name>/agent.pickle. You can also specify other agent directory by using --agents-dir parameter. Note, to use the pre-defined rllib loading routine, the agent directory name must start with 'rllib', and contain the appropriate rllib checkpoint, config, and metadata files. For more detailed info and instructions see the RllibDummy_CrampedRoom example agent.

If a more complex or custom loading routing is necessary, one can subclass the OvercookedGame class and override the get_policy method, as done in DummyOvercookedGame. Make sure the subclass is properly imported here

Saving trajectories

Trajectories from games run in overcooked-demo can be saved. By using --trajectories-dir you can specify directory that will be used to save trajectories. By default trajectories are saved inside static/assets/trajectories directory.

Replying trajectories

Trajectories from specified --trajectories-dir can be replayed in http://localhost/replay.

Updating Overcooked_ai

This repo was designed to be as flexible to changes in overcooked_ai as possible. To change the branch used, use the --overcooked-branch parameter shown above.

Changes to the JSON state representation of the game will require updating the JS graphics. At the highest level, a graphics implementation must implement the functions graphics_start, called at the start of each game, graphics_end, called at the end of each game, and drawState, called at every timestep tick. See dummy_graphcis.js for a barebones example.

The graphics file is dynamically loaded into the docker container and served to the client. Which file is loaded is determined by the --graphics parameter. For example, to server dummy_graphics.js one would run

./up.sh --graphics dummy_graphics.js

The default graphics file is currently overcooked_graphics_v2.2.js

Configuration

Basic game settings can be configured by changing the values in config.json

Legacy Code

For legacy code compatible with the Neurips2019 submission please see this branch of this repo.