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 [2]:
# 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:
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:
Agent.py file:
Epsilon (exploration factor) - A higher value of epsilon will mean the Q-learning algorithm will almost always choose a random action to explore more, whereas, a lower value of epsilon will indicate algorithm will choose the highest Q-value option (exploitation). Choosing a highest Q-value everytime will make the algorithm is doing better but it might ignore other paths that may give more optimal solution later.
Alpha (learning rate) - Alpha value of 0 will indicate no learning, that is, agent will not use any of the previous mistakes it made to be considered into next iteration. However, alpha equal to 1 will indicate the algorithm will not give weight to present situation and will predict based on past learning only.
Number of dummies - Number of dummies indicates the other cars present in the Simulation. If there are less cars in simulation, smartcab will encounter less traffic and reach destination fairly early. On the other hand, if there are too many cars in simulation, there might be a deadlock situtation where nobody can move because there is no empty space. So smartcab will not be able to reach its destination.
Environment.py file:
Simulator.py file:
Planner.py file:
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 [2]:
# 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: From the visualization above, following are the observations:
The agent consistently makes bad decisions. The agent is making these bad decisions with a total frequency of 50% of time during trials 12 to 18. Out of all those bad incidents, about 6% of the time during each trial the agent is in a accident.
Yes, the rate of reliability makes sense as the agent is driving randomly. The agent will take a random action in this case which may bring it closer to destination or take agent into opposite direction to that of destination.
The agent is receiving negative rewards for all the trials. Yes, when a major accident happen, the agent gets punished heavily by being given a significant negative reward. This is seen in rewards visualization between trials 12 and 14 when a major accident happened and rewards decreased.
No, as the trials increases the agent is experiencing more bad incidents instead of being decreased.
Given the reliability rating of F and increasing number of incidents at every trial, this version of Smartcab cannot be considered reliable or safe for passengers.
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: The features relevant for safety and efficiency of a smartcab are:
'waypoint' : A smartcab needs to know in which direction it needs to move forward which will bring it closer to destination. Leaving this feature behind will make Smartcab to go and look in all possible directions, which is not efficient.
'light' from 'inputs' - The lights feature is a necessary one since the agent should not move during red light. If light feature is not there, agent might move during red light which might result in accident due to traffic from orthognal direction.
'left' from 'inputs' - If the smartcab needs to make a right turn at red light, it needs to know if there is any vehicle coming in from left side so as to avoid collision.
'right' from 'inputs' - If the smartcab wants to take a right, it needs to make sure there is no car in right and there is space for going in right direction otherwise there will be a collision.
'oncoming' from 'inputs' - If Smartcab needs to take a left during green light, the oncoming traffic should be known. If a vehicle is coming from opposite side and going straight, Smartcab needs to wait to avoid accident.
The 'deadline' feature is not required as the deadline may be very short in which case Smartcab would make wrong decisions and get into accident.
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: The number of possible values for each feature are:
So, the total state space would be 3x2x4x4x4 = 384 states.
It would be possible for the agent to learn policy for all the 384 states only if the number of training trials are large that explore, if not all, but most of the states.
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 [10]:
# 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:
The observations that are similar to basic driving agent are that both agents start with poor reliability. Both of them fail to reach their destination in time initially. Also, both the agents are involved in high frequency of incidents (including major accidents) during initial trials and their negative rewards indicate the same.
The driving agent required approximately 25 trials. No, the number of trials are very less compared to number of states (384). The epsilon-tolerance is 0.05 and since epsilon is linear, the number of trials are very less. The agent needs more trials in order to learn from its mistakes.
Yes, the exploration factor (epsilon) is accurately represented. The epsilon is initially higher than one so that initially agent learns maximum by choosing random actions instead of choosing learned value (max Q-value). As the trials increase, the agent relies less on random actions and more on learned values (represented as linearly decreasing line).
Yes, as the training trials increased, number of bad incidents decrease significantly as agent is able to deduce correct action based on its learning and getting higher rewards for good action and negative rewards for incidents. The average reward increased in proprotion to number of trials.
The agent's safety is similar to the basic driving agent. This will be due to agent not able to learn all of the states. The agent, however, improved its reliability a little bit. The basic driving agent failed both safety and reliability ratings while default Q-learning agent receives F in safety and D in reliability.
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 [11]:
# 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:
The decaying function used for epsilon is epsilon = 1.75 e^(at), where a = 0.0075
The improved Q-learning agent took approximately 509 trials before beginning testing.
The tolerance value I used is the default one (0.05). This is to make sure the agent doesn't quit learning early since epsilon is decaying fast as number of trials increases.
The alpha value used is 0.5. This will make agent to split its focus equally on learning from action taken during this take along with what it has learned in past (old Q-value).
The improved Q-learning achieved highest ratings (safety-A+ and reliability-A+) with significantly more number of trials as compared to default Q-learner. The improved Q-learner also increases its reliability and reward at around 200th trial (at this time epsilon equals alpha value) significantly while reducing the number of bad incidents by a large margin. This improvement is likely because Q-learner agent is using the highest Q-value it has learned from past instead of randomly choosing its actions.
Yes, the Q-learner agent successfully learns optimal policy by learning on higher number of trials as compared to default Q-learner. The increasing rate of reliability while simultaneously decreasing number of bad incidents indicates that Q-learned improved significantly.
Yes, the Smartcab agent learned successfully and achieved the highest ratings (A+ for both safety and reliability)
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:
The state dictionary looks like below:
The optimal policy for the given environment would be
If there is a red light, and agent wants to go right, it can only do so if there is no traffic coming from agent's left.
If there is a green light, and agent wants to go straight, it can go freely. Agent should never stop and wait in this situation.
Since the safety and reliability of this agent is very high, many of the states converge to correct policy and give the correct action a higher Q-value than others. However, there are some states where the action having highest Q-value is not the optimal policy.
Few examples of optimal policy
('right', 'red', 'forward', 'forward', None) -- forward : -38.77 -- None : 0.35 -- right : -18.70 -- left : -30.14
Here, the smartcab wants to go 'right' on a 'red' light. There is incoming traffic on left side, traffic in right is going straight and and lastly there is no traffic in the front.
The Q-values for agent is negative for going to left or forward direction. This is correct since on a red light, agent should never proceed. Also, the agent has high penalty for going right too since there is traffic on left. The agent learned correctly to do nothing at this step as this is the best action.
('forward', 'green', 'left', 'forward', None) -- forward : 1.51 -- None : -2.16 -- right : 0.46 -- left : 0.29
Similarly in above example, the agent wants to go 'straight' on a green light. The traffic on left wants to go left, traffic on right wants to go straight ahead and there is no oncoming traffic.
The agent learns correctly that it cannot sit idle on green light (with -2.16 weight given to None action). The agent is free to go left or right since there is no traffic in either direction. However, since the agent wanted to go forward, the 'forward' action has the highest Q-value here.
('right', 'green', 'left', 'right', 'forward') -- forward : 0.37 -- None : 0.00 -- right : 0.00 -- left : 0.00
Here, the agent wants to 'right'. The signal is green for agent. The traffic on left wants to go left, while the traffic in right wants to go right. There is also oncoming traffic going straight.
The agent learned enough that it cannot sit idle and cannot go left too since there is oncoming traffic. However, agent is free to go right (which the next waypoint indicates it should do) but instead have a higher Q-value for 'forward' direction.
The agent might not have learned optimal policy either due to it did not encounter enough trials of this situation that it can learn through. The other reason can be epsilon was high at this time and so agent was forced to choose random actions in past which affected the Q-value of states and hence agent learned a different policy.
'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:
Smartcab - The smartcab is limited in its features. The agent only has information of going into the next step by use of waypoints, traffic and light status at particular intersection. The agent doesn't have full information at this point of what's happening at other intersection. The future rewards will require the agent to have knowledge of the entire grid. Hence, future rewards will not work with this agent with present features.
Environment - In this environment, the destination changes during each trial. So in one trial the destination might be located at top left corner and in the next trial diagonally opposite with agent's initial position also changing. The future rewards in one trial might learn to go towards a destination in particular direction. The future rewards will learn to increase rewards in a particular direction. The next trial will change destination and hence direction of where agent should go. This will make future rewards learned in previous trial irrelevant to this iteration.
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.