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.
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.
In [1]:
# Import the visualization code
import visuals as vs
# Pretty display for notebooks
%matplotlib inline
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.
In a few sentences, describe what you observe during the simulation when running the default agent.py agent code. Some things you could consider:
Hint: From the /smartcab/ top-level directory (where this notebook is located), run the command
'python smartcab/agent.py'
Answer:Smartcab does not move .It is at rest at same point. It receives postive points when the red light is on ,and negetive points when the light is Green and there is no incoming traffic. But if there is a incoming traffic and smartcab is idle at green light then smartcab receives positive rewards.
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.
agent.py Python file, choose three flags that can be set and explain how they change the simulation.environment.py Python file, what Environment class function is called when an agent performs an action?simulator.py Python file, what is the difference between the 'render_text()' function and the 'render()' function?planner.py Python file, will the 'next_waypoint() function consider the North-South or East-West direction first?Answer:
Below are the agent.py Python file parameteres that effect the simulation.
learning - Whether the agent is expected to learn from rewards and penalties it receives for each action,if learning =True then model will be allowed to learn.
epsilon - Random exploration factor, a probability, 0.0-1.0 - each time step the cab will take a random action with this probability. It will choose randomly from untried actions for the current state.
alpha - Learning rate, 0.0-1.0 - at alpha=0, no learning takes place - the agent does not update the Q table based on rewards/punishments. At alpha=1, the value in the Q table is replaced directly with the current reward/punishment. At values in between, the value in the Q table is a mix of existing rewards/punishments and the current reward/punishment.
environment.py:The act method is called when an agent performs an action
simulator.py:render() renders the simulation to a PyGame screen, while render_text() renders it to the console
planner.py: next_waypoint() will consider the East-West direction first
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.
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.
In [9]:
# Load the 'sim_no-learning' log file from the initial simulation results
vs.plot_trials('sim_no-learning.csv')
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:
Answer:
The driving agent make bad decision around 43% of the time.Around 6% of those are major acccidents and 1% are minor accidents.
The agent achieves success approximately 15-20% of the time. this means that the agent arrives to the destination correctly on random around 20% of the time.
The agent is receiving rewards of about -6 points per decision, so it is being penalized pretty heavily.
The outcomes of the agent doesnt change significantly with an increase in number of trials. this is expected since the agent isnt learning or changing the behaviour. the entire trial is always random so the constant results makes sense here.
This Smartcab would NOT be considered safe or reliable for its passengers - its safety rating is F , and reliability rating is F.
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!
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.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:
Waypoints and inputs are most relevant for learning both safety and efficiency - they will let the Smartcab decide if it should stay where it is, go left, go straight, or go right, without breaking any laws or causing an accident.
The color of the light determines whether the agent should move or not, if it moves when the light is red, this would be very hazardous to the other drivers, thus light is a very strong feature for safety.
oncoming feature is important because the oncoming traffic should not intersect with the direction of travel of the vehicle, which would lead to a major accident. it is also important to know this value to learn the rule of being able to take a left turn on a green light, since left turn is not allowed if the oncoming traffic is moving to the right.
inputs[left] feature is only important if the vehicle wants to take the decision to turn left, if the vehicle is moving right, this may not affect the safety of the decision. this information is also important for learning the rule of being allowed to take a left turn when there is a red light.
waypoint and deadline influences the efficiency , because if the vehicle doesnt choose decisions which is likely to get to the destination fastest, it would not be very efficient. the input factors like light, left, right and oncoming would affect the usefulness of the decision in achieving long term efficiency, taking a route which has many oncoming traffic and vehicles is likely to be less efficient than going around a longer route without traffic, just like the real world.
The only input we don't really need is right, i.e. what the car on our right is doing - the only time we need to be careful is if it is doing a U-turn, which is not modeled by this environment. So we can ignore it, and reduce our state space. Deadline is also less important, as safety should be valued above efficiency - if the cab knows it only has one minute to get somewhere, you wouldn't want it deciding that it should speed or run red lights to get there. It might be used in planning the route though - e.g. if you had enough time you might choose to take a scenic route rather than a highway. So we'll just use waypoint, light, oncoming, left.
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!
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:
Every Feature selected can have the following states.
light == 2 (red, green)
waypoint == 3 (left, right, forward)
oncoming == 4 (None, left, right, forward)
left == 4 (None, left, right, forward)
Total size of the state space is 234*4 = 96 possible states, and the size of the state-action space is 96x4 = 384 state-actions (i.e. from each state the car can take 4 actions - None, forward, left, right).
Smartcab will wander through the state-action space semi-randomly, so it could take a couple thousand time steps to cover the state-action space.
If we set the number of training trials to 100, and each trial had 20 time steps, we'd cover 2000 state-actions, which might be enough to cover all 384 state-actions.
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!
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 $\epsilon$.
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!
In [3]:
# Load the 'sim_default-learning' file from the default Q-Learning simulation
vs.plot_trials('sim_default-learning.csv')
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:
Answer:.
safety Rating is theimproved compared to before - an F-->B.
The driving agent took 20 training trials before testing, which makes sense, as the episilon-tolerance defaults to 0.05, while epsilon starts at 1.0 and decreases by 0.05 each training trial - the output shows the expected linear decay function.
The number of bad actions increased as the number of training trials increased, reaches same level as before , while the average reward increased from about -4 to -3.5.
The Reliability Rating has increased is same as the previous model.
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.
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):
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!
In [15]:
# Load the 'sim_improved-learning' file from the improved Q-Learning simulation
vs.plot_trials('sim_improved-learning.csv')
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:
Answer:
epsilon = ABS(COS(at)) with epsilon-tolerance of 0.001 and alpha of 0.01. The cosine curve seems to matter for the agent to learn the rules of the road for an A+ safety rating, and the long tail matters for the agent to eventually follow the waypoints, but this is still not completely set since the agent still only gets an A rating and not an A+ rating. The long tail to 'epsilon-tolerance' was so that it can learn all of the safety rules in the 96 possible state combinations that I outlined in answer to Question 5.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.
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:
My example of an optimal policy would be the following for a red light when the intent is to drive forward with no oncoming traffic. The state should look like this with the possible action of None, forward, left and right:
| N | state | actions | policy |
|---|---|---|---|
| 1 | {'forward', 'red', None, None} | None | optimal |
| 2 | {'forward', 'red', None, None} | forward | incorrect |
| 3 | {'forward', 'red', None, None} | left | incorrect |
| 4 | {'forward', 'red', None, None} | right | suboptimal |
In looking at the sim_improved-learning.txt, I found this policy for this same state:
forward_red_None_None
-- forward : -3.72
-- left : -3.72
-- right : 0.19
-- None : 1.35
that is the result of the re-enforcement training. The action with the highest value of 1.35 for this policy None, matches that of the correct action in the optimal policy. There has also learned a suboptimal policy right which has a lesser value of 0.19 than the optimal one which also matches that of the correct action for the suboptimal policy. This makes sense, since in the US, you are allow to make right turns on red if there is no traffic.
I searched the sim_improved-learning.txt text file to find any suboptimal policies that the final Q-Learning agent may have learn for a state and found this:
forward_red_right_right
-- forward : -0.11
-- left : 0.00
-- right : 0.00
-- None : 0.00
In this state, the final optimized Q-Learning agent would not go forward on a red light which is the correct policy setting, but may chose instead to turn left, which has the same Q-value as right or None (idle). Left on red traffic light is clearly in violation to the US traffic rules. I would imagine that this state setting is caused by the final optimized Q-Learning agent never reaching this state in training and then chosing to make a left turn that would cause it to learn that left turn on red would result in a bad action with negative rewards. Thus, without searching the full Q-Learning feature space and learning all the negative consequences from the environment, the final Q-Learning agent may not be as optimized as we would like and may actually fail at some tests passed 100 n_tests.
Another suboptimal state from the sim_improved-larning.txt text file is this:
right_green_forward_None
-- forward : 0.17
-- left : 0.06
-- right : 0.10
-- None : -0.44
In this state, the final optimized Q-Learning agent would go forward instead of turn right since that action has the highest Q-value, even though right is the waypoint to the destination in the environment. Instead right turn only has 0.10 Q-value much lower than 0.17 Q-value for forward. This means the final Q-Learning agent is not as efficient as it could be. This maybe one of the reason why the final Q-Learning agent only has an A instead of A+ rating for reliability.
'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!
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:
First reason has to do with how the Smartcab agent can only make decisions based on states derived from a generic intersection in a grid world with no precievable location information. Currently the environment does not expose this (X,Y) coordinate location information to the Smartcab agent. Thus, location information cannot be expressed in the Q-Learning agent's state table for it to bias its action against in the grid world. In other words, the Smartcab has no knowledge of its (X,Y) location in the grid world, so cannot be bias to move away from that to another (X,Y) location that it also does not know about. It can only go towards its destination based on directions given by its environment alone at its current state.
Second reason is that, for the environment, the destination goals are randomly selected after each training or testing session. This means that the agent, besides requiring current location (X,Y) coordinates, would also need to learn a constantly changing goal. This combination does not yield a good way to calculate a future reward that will lead to good driving decisions for the agent, since a future reward would be a location/destination based policy and not a traffic rule based policy. So, this means updating the Q-value calculation at time t by including future rewards will, in fact, make the Smartcab agent start learning bad actions since the negative rewards generated by the bad actions are much smaller than the accumulated future rewards of meeting the goal. Furthermore, there are no way to make the future rewards smaller as compared to that of the negative rewards without invalidating the future reward goal of bais the action of the agent. To do so would make the agent unlearn its safety policies. In any case, the randomness of the location goal would eventually, given enough trials, make every intersection have the same bias, and thus become useless for the Smartcab agent to learn anyway.
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.