Chapter 1

1. What does the acronym PID stand for? Is this considered an AI software method?

   PID stands for Proportional, Integral, Derivative, and is a type of closed-loop controller that does not require a model (simulation) to operate. PID is not an AI method because there is no learning or adaptation involved in the decision-making. PIDs are very useful control techniques and are widely used to control motors and thermostats.

2. What is the Turing Test? Do you feel that this is a valid method of assessing an artificial intelligence system?

   The Turing Test, originally called The Imitation Game by Alan Turing, is an imaginary test, or thought experiment, in which a person communicates with someone or something via a teletype (or a text message, for you younger people). An AI would pass the Turing Test if the person was unable to tell whether the entity they were communicating with was a human or a robot. The Turing Test has been passed by modern AI-based chatbots and generative AI...

Chapter 2

1. Name three types of robot sensors.

   A sensor is anything that conveys data from the outside world to the robot. Sensors mentioned in the text include the following:

   • Sonar sensors
   • Cameras
   • Microphone
   • Buttons
   • Analog-to-digital voltage sensors
   • Temperature via thermistors
2. What does the acronym PWM stand for?

   PWM stands for pulse width modulation, a type of digital-to-analog control scheme where pulses are sent out that get longer based on the amount of control desired. In other words, the pulse duty cycle (amount of time on/off) gets converted into a voltage to drive a motor. This is commonly used to control DC motors.

3. What is analog-to-digital conversion? What goes in and what comes out?

   As the name says, analog-to-digital (A2D) conversion takes in an analog value, typically a voltage, and converts it into a digital value or number that the digital part of the computer can understand. A typical application is measuring battery voltage to determine the state of charge.

...

Chapter 3

1. Describe some of the differences between a storyboard for a movie or cartoon and a storyboard for a software program.

   A storyboard for a movie is used not only for advancing the plot, but also for showing what point of view will be used – in other words, it is used to plan camera angles, directions, and movements. In that the purpose of both storyboards is to “tell the story” of what happens, they are the same. The point of view of a computer software storyboard should be the user.

2. What are the five W questions? Can you think of any more questions that would be relevant in examining a use case?

   Who, What, When, Where, Why (with Why being the most important). More relevant questions might be: How well? How often? How many or how much?

3. Complete this sentence: A use case shows what the robot does but not ________.

   “How it does it.” Use cases are from the user’s perspective and never include implementation details.

4. Take...

Chapter 4

1. We went through a lot in this chapter. You can use the framework provided to investigate the properties of neural networks. Make adjustments to the learning rate, batch size, number of epochs, and loss functions.

   This is an exercise for the student. You should see different curves develop as these parameters are changed. Some will not produce an answer at all (which looks like random results – the curve stays at the same level as no learning is taking place). Some will learn faster or slower.

2. Draw a diagram of an artificial neuron and label the parts. Look up a natural, human biological neuron and compare.

   See Figure 4.3 in the chapter. The artificial neuron has a number of inputs, a set of weights, one for each input, a bias, an activation, and a set of outputs.

3. Which features are the same in a real neuron and an artificial neuron?

   Both have multiple inputs and multiple outputs and accept inputs, perform some processing, and then make an output. Both use...

Chapter 5

1. In Q-learning, what does the Q stand for (you will have to research this on the internet)?

   The origin of Q-learning is the doctoral thesis of Christopher John Cornish Hellaby Watkins from King’s College, London, May, 1989 (https://www.researchgate.net/publication/33784417_Learning_From_Delayed_Rewards). Evidently, the Q just stands for Quantity.

2. What could we do to limit the number of states that the Q-learning algorithm must search through?

   Only pick the Q-states that are relevant and are follow-ons to the current state. If one of the states is impossible to reach from the current position, or state, then don’t consider it.

3. What effect does changing the learning rate have on the learning process?

   If the learning rate is too small, the training can take a very long time. If the learning rate is too large, the system does not learn a path but instead overshoots and may miss the minimum or optimum solution. If the learning rate is too big, the solution...

Chapter 6

1. Do some internet research on why an open source voice assistant was named Mycroft. How many stories did you find and which one did you like?

   I found at least three. My favorite is that Mycroft is Sherlock Holmes’ older, and some say smarter, brother. Sherlock Holmes is played on TV in the UK by Benedict Cumberbatch, who also played Alan Turing in the movie The Imitation Game, the original name of the Turing Test, a test of AI conversation, which is what Mycroft does.

2. In the discussion of intent, how would you design a neural network to predict command intent from natural language sentences?

   One approach would be to gather a selection of commands, label the intent of the command, use the commands as input in a neural network, and use intent as the output label for the training.

3. Rewrite the “receive knock-knock jokes” program to remember the jokes told to the robot by adding them to the joke database used by the “tell knock-knock jokes...

Chapter 7

1. Regarding SLAM, what sensor is most used to create the data that SLAM needs to make a map?

   Light detection and ranging (LiDAR) sensors are the most common SLAM sensors used by a wide margin. The 3D data that LiDAR provides is perfect for SLAM’s mapping function.

2. Why does SLAM work better with wheel odometer data available?

   The wheel odometers reduce the search space that the SLAM algorithm needs to look for the possible locations of the robot after moving. Thus, it increases information and reduces uncertainty in the map. How does it do this? By giving extra measurements about where the robot is located (how far it moved), we can then reduce our search to match where we are against our sensor readings.

3. In the Floor Finder algorithm, what does the Gaussian blur function do to improve the results?

   The Gaussian blur function reduces noise and gets rid of stray single pixels in the image, making for a smoother result.

4. The final step in Floor Finder is to...

Chapter 8

1. What are the three ways to traverse a decision tree?

   From beginning to end (start to goal); from goal to start; and from both ends at once to meet in the middle.

2. In the fishbone diagram example, how do you go about pruning the branches of the decision tree?

   By eliminating the effect of the item on a branch. For example, using our “robot does not move” fault, if the branch says “Arduino-no power” and you check to see if the Arduino has power and it does, you can prune that branch. If the branch is “motor stuck”, the effect of having a motor stuck is that the robot will drive in circles. As the robot is not driving in circles – it is not driving at all – you can prune that branch.

3. What is the role of the Gini coefficient in creating a classification?

   It determines the amount of impurity in the sample or pool. When the Gini coefficient = 0, all of the members of the class have the same attributes, and no further...

Chapter 9

1. What is your favorite movie robot? How would you describe its personality?

   This is, of course, a subjective question. I’m a big R2D2 fan. R2 is feisty, determined, and stubborn as well as being a faithful companion and helper. R2 will get you out of a jam, fix your starfighter, provide cover from hostile fire, and hack Imperial computers. He is a Swiss army knife on wheels.

2. What techniques did the movie-makers use to express R2D2’s personality (body language, sounds, etc.)?

   R2D2 owes his personality to a combination of his emotional beeps and squawks (provided by Ben Burtt) and his body movements provided by having a person inside his chassis (Kenny Baker). They were stuck with the not-very versatile chassis designed for the first Star Wars movie, which only has a head that moves. Most of R2’s persona comes through in his sounds, including his famous scream.

3. What are the two types of chatbots? List some of the strengths and weaknesses of each...

Chapter 10

1. Given that we started the chapter with knock-knock jokes and ended up talking about robot phobia and addressing philosophical questions about existence, do you feel that AI is a threat? Why or why not?

   I do not feel that robots or AI are a threat in any way because the necessary and sufficient conditions for robots to be a threat do not exist, which is to say that the robots have to want to take over the world and must have a need to take over. Currently, robots and AI have no such wants or needs.

2. List any five professions that would be necessary to turn our Albert robot into a product.

   We would need project managers, packaging designers, advertising and marketing experts, salespeople, engineers, technicians, artists, package designers, machinists, electricians, accountants, lawyers, a psychologist, and support staff, among others. You can select any five from among these options.

3. Why would we need a psychologist in our imaginary company that manufactures robots...