Our objective is to train an RL agent to optimize pathfinding strategies for guiding a pitchfork within a grid-based warehouse environment.

The RL agent learns to navigate the pitchfork target location in the minimum number of steps while avoiding obstacles and crates. The agent can move a pitchfork or a crate to guide the pitchfork to the desired location.

We tried several RL approaches to solve this like:

• Value Iteration

• SARSA

• Q-Learning

• SARSA with eligibility traces

• Q-Learning with eligibility traces