Diagonal Movement and Cycle Detection

[weihangzheng]

Overview

This update focuses on refining the agent's decision-making process in our simulation environment, with an emphasis on advanced observation handling and enhanced movement capabilities. The enhancements aim to process observation data efficiently, detect repetitive patterns, and allow for more versatile movement strategies. Key aspects include:

1. Cycle Detection: Implementing a direct comparison method to detect identical observation dictionaries without hashing, for more efficient cycle detection.
2. Enhanced Movement Capabilities: Enabling agents to move diagonally, offering more dynamic and flexible movement strategies within the simulation environment.

Changes Made

• Cycle Detection without Hashing: Added a method for detecting identical observation dictionaries directly, bypassing the need for hashing, to improve cycle detection efficiency.
• Past Observations Storage: Updated agent_instance.past_observations to be maintained as a list, allowing new observations to be appended easily for cycle detection.
• Diagonal Movement: Implemented the capability for agents to move diagonally, enhancing their ability to navigate and interact within the environment.