Machine Learning Engineer Nanodegree¶

Reinforcement Learning¶

Project: Train a Smartcab to Drive¶

Welcome to the fourth project of the Machine Learning Engineer Nanodegree! In this notebook, template code has already been provided for you to aid in your analysis of the Smartcab and your implemented learning algorithm. You will not need to modify the included code beyond what is requested. There will be questions that you must answer which relate to the project and the visualizations provided in the notebook. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide in agent.py.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode.

Getting Started¶

In this project, you will work towards constructing an optimized Q-Learning driving agent that will navigate a Smartcab through its environment towards a goal. Since the Smartcab is expected to drive passengers from one location to another, the driving agent will be evaluated on two very important metrics: Safety and Reliability. A driving agent that gets the Smartcab to its destination while running red lights or narrowly avoiding accidents would be considered unsafe. Similarly, a driving agent that frequently fails to reach the destination in time would be considered unreliable. Maximizing the driving agent's safety and reliability would ensure that Smartcabs have a permanent place in the transportation industry.

Safety and Reliability are measured using a letter-grade system as follows:

Grade	Safety	Reliability
A+	Agent commits no traffic violations, and always chooses the correct action.	Agent reaches the destination in time for 100% of trips.
A	Agent commits few minor traffic violations, such as failing to move on a green light.	Agent reaches the destination on time for at least 90% of trips.
B	Agent commits frequent minor traffic violations, such as failing to move on a green light.	Agent reaches the destination on time for at least 80% of trips.
C	Agent commits at least one major traffic violation, such as driving through a red light.	Agent reaches the destination on time for at least 70% of trips.
D	Agent causes at least one minor accident, such as turning left on green with oncoming traffic.	Agent reaches the destination on time for at least 60% of trips.
F	Agent causes at least one major accident, such as driving through a red light with cross-traffic.	Agent fails to reach the destination on time for at least 60% of trips.

To assist evaluating these important metrics, you will need to load visualization code that will be used later on in the project. Run the code cell below to import this code which is required for your analysis.

# Import the visualization code
import visuals as vs

# Pretty display for notebooks
%matplotlib inline

Understand the World¶

Before starting to work on implementing your driving agent, it's necessary to first understand the world (environment) which the Smartcab and driving agent work in. One of the major components to building a self-learning agent is understanding the characteristics about the agent, which includes how the agent operates. To begin, simply run the agent.py agent code exactly how it is -- no need to make any additions whatsoever. Let the resulting simulation run for some time to see the various working components. Note that in the visual simulation (if enabled), the white vehicle is the Smartcab.

Question 1¶

In a few sentences, describe what you observe during the simulation when running the default agent.py agent code. Some things you could consider:

Does the Smartcab move at all during the simulation?
What kind of rewards is the driving agent receiving?
How does the light changing color affect the rewards?

Hint: From the /smartcab/ top-level directory (where this notebook is located), run the command

'python smartcab/agent.py'

Answer: Description : When running the default agent.py agent code we observe that the world is a 8 by 6 grid. Each column and each row has a road resulting in 48 intersections. The agent is a white car starting at a random location in the environment and a Udacity logo shows the destination it needs to reach. The simulation also involves 100 other cars moving in the world. Traffic lights are present at every intersections. Indications are written on top of the GUI to show informations about the simulation such as the number of trial, if the agent succeeded in the previous trial, the action chose by the agent and the associated reward for that action, the time remaining to reach destination...

In the current state, the Smartcab is not moving at all during the simulation.

The driving agent is receiving positive and negative rewards.

After few minutes of observation, I see that the reward is slightly positive when "the agent is idled at red light" and strongly negative if not moving when "there was a green light with no oncoming traffic". According to the weight of each reward we can say that the agent is more penalized for doing nothing when it was supposed to move (unreliable) than rewarded for doing nothing when it was supposed to do nothing (safety). Therefore being unreliable is more penalizing here than it is rewarded to be safe.

Understand the Code¶

In addition to understanding the world, it is also necessary to understand the code itself that governs how the world, simulation, and so on operate. Attempting to create a driving agent would be difficult without having at least explored the "hidden" devices that make everything work. In the /smartcab/ top-level directory, there are two folders: /logs/ (which will be used later) and /smartcab/. Open the /smartcab/ folder and explore each Python file included, then answer the following question.

Question 2¶

In the agent.py Python file, choose three flags that can be set and explain how they change the simulation.
In the environment.py Python file, what Environment class function is called when an agent performs an action?
In the simulator.py Python file, what is the difference between the 'render_text()' function and the 'render()' function?
In the planner.py Python file, will the 'next_waypoint() function consider the North-South or East-West direction first?

Answer:

In the agent.py Python file, choose three flags that can be set and explain how they change the simulation.

In the agent.py, I count 13 different flags. I choose to explain the first three in the class Environment. The first is the boolean verbose. When set to True, additional output from the simulation are displayed such as the time step or if the agent has reached destination. Those additional information can really useful while developing the simulation. There is also num_dummies which is a discrete number of dummy agents in the environment. It is set to 100 by default. This variable can impact the reliability of the agent as it might delay the agent's progression towards the destination. It can also impact the safety as it will increase the probability of the agent to be involved in an accident. Finally the grid_size is a discrete number that defines the number of intersections. By default columns = 8 and rows = 6. Changing the size of the world can have an impact on the time it will take for the agent to reach the destination and also impact the number of interactions tha agent will have with the environment.

In the environment.py Python file, what Environment class function is called when an agent performs an action?

The function 'act()' is called when an agent performs an action. It is defined by the arguments agent and action and make the agent perform the action if it is legal.

In the simulator.py Python file, what is the difference between the 'render_text()' function and the 'render()' function?

The 'render_text ()' function, as it is defined, is the non-GUI render display of the simulation and the simulated trial data are rendered in text format in the terminal/command prompt. Whereas the 'render()' function is the GUI render display of the simulation and allows to visualize the simulation.

In the planner.py Python file, will the 'next_waypoint() function consider the North-South or East-West direction first?

The 'next_waypoint() function will first consider the East-West direction.

Implement a Basic Driving Agent¶

The first step to creating an optimized Q-Learning driving agent is getting the agent to actually take valid actions. In this case, a valid action is one of None, (do nothing) 'left' (turn left), right' (turn right), or 'forward' (go forward). For your first implementation, navigate to the 'choose_action()' agent function and make the driving agent randomly choose one of these actions. Note that you have access to several class variables that will help you write this functionality, such as 'self.learning' and 'self.valid_actions'. Once implemented, run the agent file and simulation briefly to confirm that your driving agent is taking a random action each time step.

Basic Agent Simulation Results¶

To obtain results from the initial simulation, you will need to adjust following flags:

'enforce_deadline' - Set this to True to force the driving agent to capture whether it reaches the destination in time.
'update_delay' - Set this to a small value (such as 0.01) to reduce the time between steps in each trial.
'log_metrics' - Set this to True to log the simluation results as a .csv file in /logs/.
'n_test' - Set this to '10' to perform 10 testing trials.

Optionally, you may disable to the visual simulation (which can make the trials go faster) by setting the 'display' flag to False. Flags that have been set here should be returned to their default setting when debugging. It is important that you understand what each flag does and how it affects the simulation!

Once you have successfully completed the initial simulation (there should have been 20 training trials and 10 testing trials), run the code cell below to visualize the results. Note that log files are overwritten when identical simulations are run, so be careful with what log file is being loaded! Run the agent.py file after setting the flags from projects/smartcab folder instead of projects/smartcab/smartcab.

# Load the 'sim_no-learning' log file from the initial simulation results
vs.plot_trials('sim_no-learning.csv')

Question 3¶

Using the visualization above that was produced from your initial simulation, provide an analysis and make several observations about the driving agent. Be sure that you are making at least one observation about each panel present in the visualization. Some things you could consider:

How frequently is the driving agent making bad decisions? How many of those bad decisions cause accidents?
Given that the agent is driving randomly, does the rate of reliability make sense?
What kind of rewards is the agent receiving for its actions? Do the rewards suggest it has been penalized heavily?
As the number of trials increases, does the outcome of results change significantly?
Would this Smartcab be considered safe and/or reliable for its passengers? Why or why not?

Answer:

In the '10-Trial Rolling Relative Frequency of Bad Actions' we can see that the relative frequency of bad actions over 10 trials is comprised between 0.4302 and 0.4840. The relative frequency of major accident is approximately 0.055 and for minor accidents approximately 0.025 (in average). We can then consider that around 18% of the bad decisions cause accidents.
In the '10-Trial Rolling Rate Reliability' we see that the rate of reliability is comprised between 0 and 20% which is very low. Considering that all actions are chosen randomly, this rate of reliability makes sense.
In the '10-Trial Rolling Average Reward Per Action' we can see that the average reward per action are mostly negative around -5. As we can see in environment.py def act() function, the rewards are -5 for minor violations, -10 for major violations, -20 for minor accidents, -40 for major accidents. This shows that in average, the agent made a minor violation per action and suggest that it has been heavily penalized for moving randomly.
In the '10-Trial Rolling Rate Reliability' we can see the the outcome does not change significantly as the number of trials increases and stays between 0 and 20%. Also, the '10-Trial Rolling Average Reward Per Action' shows a very consistent -5 rewards per action showing no improvement whatsoever. This observation makes sense considering that all decision are made randomly.
This Smartcab is far from being safe and/or reliable as we can see by the notations, Safety rating : F and Reliability rating : F. The F notation for the reliability rating means that the success ratio < 0.6 (as coded in visual.py 'calculate_reliability()' function). The F notation for the safety rating means that the agent has been involved in at least one major accident (as coded in the visual.py 'calculate_safety()' function). Therefore the smartcab cannot be considered safe because it's prone to accidents nor reliable because it gets to its destination on time less than 60 % of the time.

Inform the Driving Agent¶

The second step to creating an optimized Q-learning driving agent is defining a set of states that the agent can occupy in the environment. Depending on the input, sensory data, and additional variables available to the driving agent, a set of states can be defined for the agent so that it can eventually learn what action it should take when occupying a state. The condition of 'if state then action' for each state is called a policy, and is ultimately what the driving agent is expected to learn. Without defining states, the driving agent would never understand which action is most optimal -- or even what environmental variables and conditions it cares about!

Identify States¶

Inspecting the 'build_state()' agent function shows that the driving agent is given the following data from the environment:

'waypoint', which is the direction the Smartcab should drive leading to the destination, relative to the Smartcab's heading.
'inputs', which is the sensor data from the Smartcab. It includes
- 'light', the color of the light.
- 'left', the intended direction of travel for a vehicle to the Smartcab's left. Returns None if no vehicle is present.
- 'right', the intended direction of travel for a vehicle to the Smartcab's right. Returns None if no vehicle is present.
- 'oncoming', the intended direction of travel for a vehicle across the intersection from the Smartcab. Returns None if no vehicle is present.
'deadline', which is the number of actions remaining for the Smartcab to reach the destination before running out of time.

Question 4¶

Which features available to the agent are most relevant for learning both safety and efficiency? Why are these features appropriate for modeling the Smartcab in the environment? If you did not choose some features, why are those features not appropriate? Please note that whatever features you eventually choose for your agent's state, must be argued for here. That is: your code in agent.py should reflect the features chosen in this answer.

NOTE: You are not allowed to engineer new features for the smartcab.

Answer:

The next waypoint, the intersection inputs, and the deadline are all features available to the agent. For learning both safety and efficiency, the most relevant features are:

'waypoint' because it shows the direction to the agent which is primordial for efficiency
'light', 'left', 'oncoming' because it let's the agent know is there are vehicules around before going a certain direction and traffic lights so it's important for safety.

I don't have a driving license and we don't have this rule in France so I might be wrong on this one but I didn't include the 'right' because as far as I understand the U.S. Right-of-Way rules, the agent will not need to yield to traffic from the right side of the intersection when it is making any move. This is assuming that the dummy agents all follow traffic laws and right traffic will not run through a red light. Also in this environment, the agent can turn right at a red light as long as it avoids collision with left traffic moving forward.

Also I didn't include 'deadline' because giving too much important on the time left to reach a destination might lead the agent to make violations or worse get into an accident. Since 'waypoint' is already supposed to give the best way, then it should be the shortest way and what we care about is to get there as soon as possible given that we are following the rules. It's good to get somewhere on time, it's always better alive.

Define a State Space¶

When defining a set of states that the agent can occupy, it is necessary to consider the size of the state space. That is to say, if you expect the driving agent to learn a policy for each state, you would need to have an optimal action for every state the agent can occupy. If the number of all possible states is very large, it might be the case that the driving agent never learns what to do in some states, which can lead to uninformed decisions. For example, consider a case where the following features are used to define the state of the Smartcab:

('is_raining', 'is_foggy', 'is_red_light', 'turn_left', 'no_traffic', 'previous_turn_left', 'time_of_day').

How frequently would the agent occupy a state like (False, True, True, True, False, False, '3AM')? Without a near-infinite amount of time for training, it's doubtful the agent would ever learn the proper action!

Question 5¶

If a state is defined using the features you've selected from Question 4, what would be the size of the state space? Given what you know about the environment and how it is simulated, do you think the driving agent could learn a policy for each possible state within a reasonable number of training trials?
Hint: Consider the combinations of features to calculate the total number of states!

Answer:

I chose the features 'waypoint', 'light', 'left' and 'oncoming'. For each features we have :

'waypoint' 3 possible outcome : 'forward', 'right', 'left'
'light' 2 possible outcome : 'green', 'red'
'left' 4 possible outcome : 'None', 'forward', 'right', 'left'
'oncoming' 4 possible outcome : 'None', 'forward', 'right', 'left'

Therefore size of state space $= 3 \times 2 \times 4 \times 4 = 96$

A size of state space of 96 is still fairly reasonable considering all the other features we could take into account. The environment has been extensively simplified and I think that the driving agent will be able to learn a policy for each possible state within a reasonable number of training trials.

Update the Driving Agent State¶

For your second implementation, navigate to the 'build_state()' agent function. With the justification you've provided in Question 4, you will now set the 'state' variable to a tuple of all the features necessary for Q-Learning. Confirm your driving agent is updating its state by running the agent file and simulation briefly and note whether the state is displaying. If the visual simulation is used, confirm that the updated state corresponds with what is seen in the simulation.

Note: Remember to reset simulation flags to their default setting when making this observation!

Implement a Q-Learning Driving Agent¶

The third step to creating an optimized Q-Learning agent is to begin implementing the functionality of Q-Learning itself. The concept of Q-Learning is fairly straightforward: For every state the agent visits, create an entry in the Q-table for all state-action pairs available. Then, when the agent encounters a state and performs an action, update the Q-value associated with that state-action pair based on the reward received and the iterative update rule implemented. Of course, additional benefits come from Q-Learning, such that we can have the agent choose the best action for each state based on the Q-values of each state-action pair possible. For this project, you will be implementing a decaying, $\epsilon$-greedy Q-learning algorithm with no discount factor. Follow the implementation instructions under each TODO in the agent functions.

Note that the agent attribute self.Q is a dictionary: This is how the Q-table will be formed. Each state will be a key of the self.Q dictionary, and each value will then be another dictionary that holds the action and Q-value. Here is an example:

{ 'state-1': { 
    'action-1' : Qvalue-1,
    'action-2' : Qvalue-2,
     ...
   },
  'state-2': {
    'action-1' : Qvalue-1,
     ...
   },
   ...
}

Furthermore, note that you are expected to use a decaying $\epsilon$ (exploration) factor. Hence, as the number of trials increases, $\epsilon$ should decrease towards 0. This is because the agent is expected to learn from its behavior and begin acting on its learned behavior. Additionally, The agent will be tested on what it has learned after $\epsilon$ has passed a certain threshold (the default threshold is 0.05). For the initial Q-Learning implementation, you will be implementing a linear decaying function for $\epsilon$.

Q-Learning Simulation Results¶

To obtain results from the initial Q-Learning implementation, you will need to adjust the following flags and setup:

'enforce_deadline' - Set this to True to force the driving agent to capture whether it reaches the destination in time.
'update_delay' - Set this to a small value (such as 0.01) to reduce the time between steps in each trial.
'log_metrics' - Set this to True to log the simluation results as a .csv file and the Q-table as a .txt file in /logs/.
'n_test' - Set this to '10' to perform 10 testing trials.
'learning' - Set this to 'True' to tell the driving agent to use your Q-Learning implementation.

In addition, use the following decay function for $\epsilon$:

$$ \epsilon_{t+1} = \epsilon_{t} - 0.05, \hspace{10px}\textrm{for trial number } t$$

If you have difficulty getting your implementation to work, try setting the 'verbose' flag to True to help debug. Flags that have been set here should be returned to their default setting when debugging. It is important that you understand what each flag does and how it affects the simulation!

Once you have successfully completed the initial Q-Learning simulation, run the code cell below to visualize the results. Note that log files are overwritten when identical simulations are run, so be careful with what log file is being loaded!

# Load the 'sim_default-learning' file from the default Q-Learning simulation
vs.plot_trials('sim_default-learning.csv')

Question 6¶

Using the visualization above that was produced from your default Q-Learning simulation, provide an analysis and make observations about the driving agent like in Question 3. Note that the simulation should have also produced the Q-table in a text file which can help you make observations about the agent's learning. Some additional things you could consider:

Are there any observations that are similar between the basic driving agent and the default Q-Learning agent?
Approximately how many training trials did the driving agent require before testing? Does that number make sense given the epsilon-tolerance?
Is the decaying function you implemented for $\epsilon$ (the exploration factor) accurately represented in the parameters panel?
As the number of training trials increased, did the number of bad actions decrease? Did the average reward increase?
How does the safety and reliability rating compare to the initial driving agent?

Answer:

Looking at the notation we can see that the safety rating is the same between the basic driving agent and the default Q-Learning agent. Besides this part the basic driving agents and the default Q-Learning agent do not share more similarities.
As we can see in the 'sim_default-learning.csv', it took 20 training trials before testing. As asked, I used the following decay function for $\epsilon$:

$$ \epsilon_{t+1} = \epsilon_{t} - 0.05 \hspace{10px}\textrm{for trial number } t$$

And the $\epsilon_{tolerance}$ was 0.05. As the initial $\epsilon$ value was 1, the number of trials was then $\frac{\epsilon_{t}-\epsilon_{tolerance}}{0.05} = \frac{1-0.05}{0.05} = 19$. We can see in the 'sim_default-learning.csv' file that at trial number 19, $\epsilon = 0.049999999999999684$ and it took one more trial before starting testing maybe because of value approximation.

We see that the $\epsilon$ is accurately represented as a line with a slope of -0.05 as implemented above.
In the '10-Trial Rolling Relative Frequency of Bad Actions' we can see a significant decrease of the number of total bad actions as the number of training trails increases. When looking at the '10-Trial Rolling Average Reward per Action', we can see that the reward per action increases with the number of training trails. All those observations show that the agent is learning.
As mentionned above, the safety rating for both driving agent are equivalent with an F. This means that in both cases the agent has been involved in at least one major accident. However we observe that the reliability rating has significantly improve from F to A. This means that the success ratio is $\ge$ 0.90.

Improve the Q-Learning Driving Agent¶

The third step to creating an optimized Q-Learning agent is to perform the optimization! Now that the Q-Learning algorithm is implemented and the driving agent is successfully learning, it's necessary to tune settings and adjust learning paramaters so the driving agent learns both safety and efficiency. Typically this step will require a lot of trial and error, as some settings will invariably make the learning worse. One thing to keep in mind is the act of learning itself and the time that this takes: In theory, we could allow the agent to learn for an incredibly long amount of time; however, another goal of Q-Learning is to transition from experimenting with unlearned behavior to acting on learned behavior. For example, always allowing the agent to perform a random action during training (if $\epsilon = 1$ and never decays) will certainly make it learn, but never let it act. When improving on your Q-Learning implementation, consider the implications it creates and whether it is logistically sensible to make a particular adjustment.

Improved Q-Learning Simulation Results¶

To obtain results from the initial Q-Learning implementation, you will need to adjust the following flags and setup:

'enforce_deadline' - Set this to True to force the driving agent to capture whether it reaches the destination in time.
'update_delay' - Set this to a small value (such as 0.01) to reduce the time between steps in each trial.
'log_metrics' - Set this to True to log the simluation results as a .csv file and the Q-table as a .txt file in /logs/.
'learning' - Set this to 'True' to tell the driving agent to use your Q-Learning implementation.
'optimized' - Set this to 'True' to tell the driving agent you are performing an optimized version of the Q-Learning implementation.

Additional flags that can be adjusted as part of optimizing the Q-Learning agent:

'n_test' - Set this to some positive number (previously 10) to perform that many testing trials.
'alpha' - Set this to a real number between 0 - 1 to adjust the learning rate of the Q-Learning algorithm.
'epsilon' - Set this to a real number between 0 - 1 to adjust the starting exploration factor of the Q-Learning algorithm.
'tolerance' - set this to some small value larger than 0 (default was 0.05) to set the epsilon threshold for testing.

Furthermore, use a decaying function of your choice for $\epsilon$ (the exploration factor). Note that whichever function you use, it must decay to 'tolerance' at a reasonable rate. The Q-Learning agent will not begin testing until this occurs. Some example decaying functions (for $t$, the number of trials):

$$ \epsilon = a^t, \textrm{for } 0 < a < 1 \hspace{50px}\epsilon = \frac{1}{t^2}\hspace{50px}\epsilon = e^{-at}, \textrm{for } 0 < a < 1 \hspace{50px} \epsilon = \cos(at), \textrm{for } 0 < a < 1$$ You may also use a decaying function for $\alpha$ (the learning rate) if you so choose, however this is typically less common. If you do so, be sure that it adheres to the inequality $0 \leq \alpha \leq 1$.

If you have difficulty getting your implementation to work, try setting the 'verbose' flag to True to help debug. Flags that have been set here should be returned to their default setting when debugging. It is important that you understand what each flag does and how it affects the simulation!

Once you have successfully completed the improved Q-Learning simulation, run the code cell below to visualize the results. Note that log files are overwritten when identical simulations are run, so be careful with what log file is being loaded!

# Load the 'sim_improved-learning' file from the improved Q-Learning simulation
vs.plot_trials('sim_improved-learning.csv')

Question 7¶

Using the visualization above that was produced from your improved Q-Learning simulation, provide a final analysis and make observations about the improved driving agent like in Question 6. Questions you should answer:

What decaying function was used for epsilon (the exploration factor)?
Approximately how many training trials were needed for your agent before begining testing?
What epsilon-tolerance and alpha (learning rate) did you use? Why did you use them?
How much improvement was made with this Q-Learner when compared to the default Q-Learner from the previous section?
Would you say that the Q-Learner results show that your driving agent successfully learned an appropriate policy?
Are you satisfied with the safety and reliability ratings of the Smartcab?

Answer:

I used the exponential decay for the exploration factor $\epsilon = e^{-at}, \textrm{for } 0 < a < 1 $ and I choose a value of $a = 0.01$
It took 300 training trials to the aget before testing. This make sense because knowing that $\epsilon_{tolerance} = 0.05$ and the initial $\epsilon_t = 1$
$$e^{-0.01\times t} = 0.05$$ $$0.01 \times t = 3$$ $$t = 300$$
I used a $\epsilon_{tolerance} = 0.05$ and $\alpha = 0.5$. I choose 0.5 for the learning rate because in this setup, I think it is important to learn as much from previous experiences as it is to take futur experiences into account. I didn't want to give more importance to one of them and as I didn't choose to decay the learning rate over time it was for me the best compromise. I used a $\epsilon_{tolerance}$ of 0.05 because I figured it was ideal to go through enough iteration (300) according to me.
Several improvements can be observed with the Q-Learner compared to the default Q-Learner. First, we've finally reached a safety rating of A+ which means the ratio = 1 and therefore shows a perfect driving. It as a big improvement compared to the safety rating F we got from the default Q-Learner. We also reached a reliability rating of A+ which means the success ratio = 1 and the agent always met the deadline. This trend can also be observed in the '10-Trial Rolling Rate of Reliability' where we see that after 260 trials, the agent reaches 100% rate of reliability. The number of total bad actions has gone down to almost 0 in the '10-Trial Rolling Relative Frequency of Bad Actions' when it was still around 0.16 with the previous default Q-Learner. We also see the increase in of reward per action going almost to +2 in the '10-Trial Rolling Average Reward per Action' when it got around -1 with the previous default Q-Learner.
All those observation show that the driving agent successfully learned an appropriate policy in this environment.
I am satified with the safety and reliability ratings of the smartcab.

Define an Optimal Policy¶

Sometimes, the answer to the important question "what am I trying to get my agent to learn?" only has a theoretical answer and cannot be concretely described. Here, however, you can concretely define what it is the agent is trying to learn, and that is the U.S. right-of-way traffic laws. Since these laws are known information, you can further define, for each state the Smartcab is occupying, the optimal action for the driving agent based on these laws. In that case, we call the set of optimal state-action pairs an optimal policy. Hence, unlike some theoretical answers, it is clear whether the agent is acting "incorrectly" not only by the reward (penalty) it receives, but also by pure observation. If the agent drives through a red light, we both see it receive a negative reward but also know that it is not the correct behavior. This can be used to your advantage for verifying whether the policy your driving agent has learned is the correct one, or if it is a suboptimal policy.

Question 8¶

Please summarize what the optimal policy is for the smartcab in the given environment. What would be the best set of instructions possible given what we know about the environment? You can explain with words or a table, but you should thoroughly discuss the optimal policy.
Next, investigate the 'sim_improved-learning.txt' text file to see the results of your improved Q-Learning algorithm. For each state that has been recorded from the simulation, is the policy (the action with the highest value) correct for the given state? Are there any states where the policy is different than what would be expected from an optimal policy?
Provide a few examples from your recorded Q-table which demonstrate that your smartcab learned the optimal policy. Explain why these entries demonstrate the optimal policy.
Try to find at least one entry where the smartcab did not learn the optimal policy. Discuss why your cab may have not learned the correct policy for the given state.

Be sure to document your state dictionary below, it should be easy for the reader to understand what each state represents.

Answer:

The optimal policy would be

  if Waypoint is right:
      if light green:
          then go right
      else:
          if car from left is going forward:
              then wait
          else: go right

  elif Waypoint is left:
      if light green:
          if oncoming car is going their right or forward:
              then wait
          else : go left
      else : wait

  elif Waypoint is forward:
      if light green:
          go forward
      else : wait

Most of the time the policy is correct in the 'sim_improved-learning.txt. However not all policies in the 'sim_improved-learning.txt' make sense.

Here are a few examples which demonstrate that the smartcab learned the optimal policy:

Waypoint is right, with no car from the agent's left so it turns right which is the best choice.

    ('right', 'green', 'forward', None)
     -- forward : -0.01
     -- right : 1.91
     -- None : -4.48
     -- left : -17.71

Waypoint is forward, the light is green and the oncoming car wants to turn its right. Therefore going forward is the best option.

    ('forward', 'green', 'right', 'left')
     -- forward : 1.84
     -- right : -0.03
     -- None : -2.24
     -- left : -10.17

Waypoint is right, the light is red it can turn right unless the car on the left is going forward but it's turning right so the agent can turn right.

    ('right', 'red', None, 'right')
     -- forward : -9.28
     -- right : 1.41
     -- None : 0.00
     -- left : -9.48

Here is an example where the policy is different from I expected:

The waypoint is forward and as the light is green it means the car on the left isn't moving and the oncoming car is going forward so technically nothing prevented the agent from going forward but it decided to go right instead.

     ('forward', 'green', 'forward', 'forward')
     -- forward : 0.00
     -- right : 0.53
     -- None : -3.69
     -- left : -20.02

This example shows that the agent might not have learn the right policy for this state yet because the Q-value for the right choice is 0.00 when the Q-value for right is 0.53.

Optional: Future Rewards - Discount Factor, `'gamma'`¶

Curiously, as part of the Q-Learning algorithm, you were asked to not use the discount factor, 'gamma' in the implementation. Including future rewards in the algorithm is used to aid in propagating positive rewards backwards from a future state to the current state. Essentially, if the driving agent is given the option to make several actions to arrive at different states, including future rewards will bias the agent towards states that could provide even more rewards. An example of this would be the driving agent moving towards a goal: With all actions and rewards equal, moving towards the goal would theoretically yield better rewards if there is an additional reward for reaching the goal. However, even though in this project, the driving agent is trying to reach a destination in the allotted time, including future rewards will not benefit the agent. In fact, if the agent were given many trials to learn, it could negatively affect Q-values!

Optional Question 9¶

There are two characteristics about the project that invalidate the use of future rewards in the Q-Learning algorithm. One characteristic has to do with the Smartcab itself, and the other has to do with the environment. Can you figure out what they are and why future rewards won't work for this project?

Answer:

The world is not static, there are different states for traffic lights and their are other moving cars in the environment. Also it is stochastic, futur rewards cannot be clearly determined because of all the parameters moving. The only predictable feature is the waypoint. Therefore futur rewards wouldn't work with this environment.

As for the smartcab, the most important thing for it is to learn how to drive safely. Those rules are unchanged no matter what state you're in therefore, taking futur reward into account wouldn't make any difference. Also, the agent cannot see beyond the current intersection and therefore is unable to select the optimal action in advance.

Note: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to
File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

Machine Learning Engineer Nanodegree¶

Reinforcement Learning¶

Project: Train a Smartcab to Drive¶

Getting Started¶

Understand the World¶

Question 1¶

Understand the Code¶

Question 2¶

Implement a Basic Driving Agent¶

Basic Agent Simulation Results¶

Question 3¶

Inform the Driving Agent¶

Identify States¶

Question 4¶

Define a State Space¶

Question 5¶

Update the Driving Agent State¶

Implement a Q-Learning Driving Agent¶

Q-Learning Simulation Results¶

Question 6¶

Improve the Q-Learning Driving Agent¶

Improved Q-Learning Simulation Results¶

Question 7¶

Define an Optimal Policy¶

Question 8¶

Optional: Future Rewards - Discount Factor, 'gamma'¶

Optional Question 9¶

Optional: Future Rewards - Discount Factor, `'gamma'`¶