The gym environment now includes logic to randomly position counters within the environment when it is initialized. The implementation involves creating a list of the counters, shuffling their positions and setting up a new vectorization of the environment state. This new feature contributes to a more dynamic and unpredictable simulation for testing agents' performance.

The game environment now supports vectorized states for reinforcement learning agents. This includes updates on various classes such as OrderManager, Counter, and Player. Several new utility functions are added to facilitate the conversion of the game state to vector form. The new vector state includes players, counters, orders, and game status.

The scoring system and order management have been significantly updated. The score calculation has been moved out from 'OrderManager' to external 'ScoreViaHooks' class which works via hooks. The order management updates include separating scoring from orders and removing unused functions. Changes in the configuration file and other related files updated to accommodate these modifications.

This commit modifies a key behavior in the Overcooked simulator. The flag to prevent players from squeezing into each other's positions was set to false, allowing such movements. This might affect gameplay mechanics in situations where players concurrently enter the same location.

Fixed: Burn no longer continues after taking things out of equipment, can take things out of cooking equipment again

The Overcooked environment module has been refactored to better handle player movements. The previously existing detect_collision function and its helper functions have been removed as they were no longer needed. Similarly, all the player speed values have been updated to reference a singular player_movement_speed attribute. The changes aimed to improve code readability, enforce unified player speeds, and remove redundant collision detection.

The `random_agent.py` has been refactored to use asyncio. This modification allows for more efficient handling of asynchronous operations. Additionally, a fix has been implemented to correctly identify and handle the "current nearest counter ID", making sure that the agent continues performing its tasks in a correct manner.