Week 4

Question 1

What is the function of sensors in robots?

Answer 1

Sensors allow robots to perceive their environment, enabling interaction and responsiveness to changes.

Question 2

What are some common sensors found in robots?

Answer 2

Typical sensors include sonar sensors, shaft decoders, and cameras. These provide robots with various modalities of environmental perception.

Question 3

What type of sensors are typically used in humanoid robots?

Answer 3

Humanoid robots often have force and torque sensors in their effectors to facilitate nuanced manipulation and interaction with objects.

Question 4

What are effectors in robots?

Answer 4

Effectors are the components that allow robots to move and change their body configuration.

Question 5

What does ‘degrees of freedom’ refer to in robotics?

Answer 5

Degrees of freedom refer to the independent parameters that define a robot’s configuration, determining how it can move and orient itself in space.

Question 6

Can you give an example of degrees of freedom in robots?

Answer 6

A drone has six degrees of freedom: three for its location in space (x, y, z) and three for its angular orientation (yaw, roll, pitch).

Question 7

What can a robot with a 1-DOF end effector do?

Answer 7

A robot with a 1-DOF end effector is capable of performing simple tasks such as grabbing objects.

Question 8

How does the DOF of a human hand compare to simple robotic end effectors?

Answer 8

The human hand has 27 DOF, allowing for a wide range of complex and nuanced movements.

Question 9

What marks the beginning of industrial robots?

Answer 9

The first robots in industry, like the ~1959 Unimate, were used for pre-programmed tasks such as assembling cars on a production line.

Question 10

How were early industrial robots controlled?

Answer 10

They used feedforward controllers, executing predefined sequences without feedback from the environment.

Question 11

What is a feedforward controller?

Answer 11

A type of controller that operates robots based on preset instructions without using sensory feedback to adjust its actions.

Question 12

What is an advantage of using feedforward controllers in robotics?

Answer 12

Feedforward controllers are very fast since they do not require processing feedback from the environment.

Question 13

What is a disadvantage of using feedforward controllers in robotics?

Answer 13

They are non-adaptive and cannot handle slight disturbances or changes in the environment, limiting their flexibility.

Question 14

give an example of this feedforward system

Answer 14

Question 15

What do feedback controllers do in the context of control systems?

Answer 15

Feedback controllers use information from the environment to adjust their actions.

Question 16

give an example of this feedforward system in robotics

Answer 16

keep measuring the distance until the arm reachess the object. (using the environment)

Question 17

What are the advantages of feedback control in control systems?

Answer 17

They are adaptive and robust to perturbations.

Question 18

What are the disadvantages of feedback control compared to feedforward control?

Answer 18

Feedback control is slower than feedforward control and is dependent on the quality of sensors.

Question 19

Why is it often impossible to fully understand psychological mechanisms through observation?

Answer 19

This is due to the inverse problem, which states that different mechanisms can exhibit identical behavior, making it fundamentally impossible to infer the exact mechanisms solely from behavior.

Question 20

What common error do we make when analyzing a psychological mechanism?

Answer 20

We tend to overestimate the complexity of the mechanism when analyzing it.

Question 21

What alternative method is suggested for understanding human behavior?

Answer 21

The alternative method is to work the other way around by building humans (i.e., robots with human-like behavior) instead of just analyzing them.

Question 22

what is synthetic psychology?

Answer 22

It’s centered on the idea that we can gain a deeper understanding of ourselves by creating physical models, particularly in the form of robots that mimic human behavior and processes. This field examines how combining simple components or processes can lead to complex behavior when these systems are embedded within an environment.

Question 23

What are ANNs called when they are used to control robots?

Answer 23

When ANNs are used to control robots, they are known as neural controllers.

Question 24

How are artificial neural networks (ANNs) used in robots?

Answer 24

ANNs are implemented in robots to simulate a nervous system, allowing for complex behaviors and decision-making.

Question 25

What types of neurons does an ANN-controlled robot have?

Answer 25

An ANN-controlled robot has sensory, internal, and motor neurons.

Question 26

How do neurons in an ANN transmit activation?

Answer 26

Neurons propagate activation through connections with a specific connection weight that determines the strength and influence of the signal.

Question 27

How are connection weights in an ANN determined?

Answer 27

Connection weights are usually determined through a process of training the ANN, where the network learns from experience by adjusting the weights to minimize error in its output.

Question 28

What is a **fitness function** in the context of genetic algorithms?

Answer 28

A fitness function evaluates how well an individual performs, considering factors like lifespan, food intake, and reproduction rate.

Question 29

How do **genetic algorithms with pseudocode** work?

Answer 29

1. pick a fitness function. 2. let a 100 robots perform a task. 3. keep the best 10 en remove the worst 90 4. copy the code of these 10 and at random small variations in the code. 5. repeat steps 2 till 4.

Question 30

What is **motor babbling** in the context of robot learning?

Answer 30

Motor babbling, based on Hebbian learning principles, involves robots engaging in random movements to learn and improve their motor skills.

Question 31

What is the human parallel to robotic motor babbling?

Answer 31

In humans, babies learn how to move by engaging in random movements initially, which is similar to the concept of motor babbling in robots.

Question 32

How do random movements (Motor, M) and sensory feedback (Kinesthetic, K) interact in motor babbling?

Answer 32

Random movements lead to changes in body pose and the environment, which are registered by kinesthetic cells (K), leading to frequent co-activation of M and K and the development of bidirectional associations.

Question 33

What is reinforcement learning in robotics?

Answer 33

Reinforcement learning is a type of machine learning where robots learn to make decisions by performing actions and receiving feedback in the form of rewards or punishments.

Question 34

What is the purpose of motor babbling in robots?

Answer 34

Motor babbling allows a robot to develop an internal model of its body and environment through random movements.

Question 35

How do robots use kinesthetic information in motor babbling?

Answer 35

Robots parameterize incoming kinesthetic information during motor babbling to build a model of the required movement.

Question 36

What is a reinforcement signal in robot learning?

Answer 36

A reinforcement signal is feedback sent to the robot's control system to inform it of the success or failure of its actions.

Question 37

How is the reinforcement signal typically presented in reinforcement learning?

Answer 37

The reinforcement signal is often binary in the form of pass/fail, but more complex numeric evaluations can also be used.