Use Q-learning to train a smart taxi to dirve in the best way and obeyed on the traffic lights

Get a best score 99% to arrive the destination finally

According to the principle of Q-learning and Hidden Markov, using reinforcement learning algorithms, I have built a self-learning smart taxi agent obeying traffic rules and doing automatic route planning itself.

After the strengthen training and some grid_search optimization , the agent can find the best route and arrive in time mostly. After some optimization , the success rate raised to 99%.