I need help developing the "windy gridworld" with reinfforcement learning. How does sarsa and q-learning apply to gridworld?

Prompt:

The grid will have arrows that will push an agent up when it moves onto them ( have numbers at the bottom of each column that would indicate the force of the wind). Have S be the start state and G as the goal state. The idea is for the agent to learn to get to the goal from the start in the minimum amount of steps. Formulate this as a reinforcement learning problem where each move is given a -1 value. Solv using both sarsa and q-learning. Be able to produce a graph showing the total cost of an episode throughout the training run. Preferred program would be python.