Intelligent Agents Questions and Answers

Q: Why is driving described as a continuous task environment?

The state includes positions and velocities that take any real value, and the steering and throttle inputs are likewise continuous, so no finite enumeration is possible. That forces either discretisation or continuous optimisation. Journey continuity, physical contact and changing conditions describe other properties, the last being dynamism. Identifying which variables are continuous determines which parts of the problem need which class of algorithm.

Q: What is a known environment?

Knownness is a property of the agent's knowledge rather than of the environment itself: the agent knows the transition and sensor models, though the environment may still be stochastic and partially observable. Prior exploration is one way to acquire that knowledge, observability is a separate dimension, and reliable action is determinism. Distinguishing what is true of the world from what the agent knows about it is a distinction the taxonomy is careful to preserve.

Q: What must an agent do in an unknown environment?

Without a model the agent must explore, observe what its actions do and build a model as it goes, which is exactly the reinforcement learning setting. Waiting is not an option in an operating system, optimistic assumptions are unsafe, and confining action to certain outcomes is usually impossible. An environment that must be learned while it is acted in combines the exploration problem with the control problem, which is why it is the hardest of the standard cases.

Q: Which of these is a percept for a taxi driving agent?

A percept is raw input from a sensor, and the camera image is exactly that. The lane change decision and the selected route are outputs of the agent's own reasoning, and the destination is arguably a percept when spoken but becomes a goal once adopted. Distinguishing what arrives from the environment from what the agent produces internally is the first discipline the framework imposes, and it is easy to blur in a system with many components.

Q: Which of these is an actuator for a taxi driving agent?

Actuators are whatever the agent uses to affect the world, so steering, throttle, brake, indicators and the display or voice used to speak to the passenger all qualify. Cameras are sensors, and the map and the planner are internal resources. Listing the sensors and actuators explicitly is part of the PEAS description, and doing so often reveals that the agent has less influence over its environment than the design assumed.

Q: What is a stigmergic interaction between agents?

Ants coordinate by depositing pheromone that alters the environment other ants perceive, so complex collective behaviour emerges with no messages and no central control. Swarm robotics and ant colony optimisation borrow the idea directly. Messaging, central control and imitation are alternative coordination mechanisms. Using the environment itself as the communication channel is cheap, robust and scales to very large numbers, which is why it recurs wherever many simple agents must act together.

Q: What is emergent behaviour in a multi-agent system?

Flocking, foraging and traffic patterns arise from simple local rules interacting, and the collective pattern is nowhere written down in any agent's program. It may be desirable or not. Delay, undesirability and individual dominance are incidental. Emergence is attractive because it achieves complex outcomes cheaply and troubling because it is hard to predict or guarantee, and both aspects follow from the same absence of explicit specification.

Q: Why is emergent behaviour difficult to guarantee?

Since no component specifies the collective outcome, there is nothing to check a component against, and the behaviour must be established by simulation or by analysis of the dynamics. That is a genuine obstacle in safety critical applications. Scale, isolated testing and randomness are practical complications rather than the fundamental one. A property that exists only at the level of the whole cannot be verified by examining the parts, which is a general difficulty with emergent design.

81.

Why is driving described as a continuous task environment?

A.Position, speed and steering all vary over a continuum
B.The journey continues without interruption from start to finish
C.The vehicle is in continuous contact with the road surface
D.Traffic conditions change continuously throughout the journey

Answer: A

The state includes positions and velocities that take any real value, and the steering and throttle inputs are likewise continuous, so no finite enumeration is possible. That forces either discretisation or continuous optimisation. Journey continuity, physical contact and changing conditions describe other properties, the last being dynamism. Identifying which variables are continuous determines which parts of the problem need which class of algorithm.

Discuss in Forum

82.

What is a known environment?

A.One in which the agent's actions always succeed as intended
B.One whose rules and outcomes the designer already understands
C.One that the agent has previously explored completely
D.One that is fully observable to the agent's sensors

Answer: B

Knownness is a property of the agent's knowledge rather than of the environment itself: the agent knows the transition and sensor models, though the environment may still be stochastic and partially observable. Prior exploration is one way to acquire that knowledge, observability is a separate dimension, and reliable action is determinism. Distinguishing what is true of the world from what the agent knows about it is a distinction the taxonomy is careful to preserve.

Discuss in Forum

83.

What must an agent do in an unknown environment?

A.Assume the environment is deterministic and fully observable
B.Restrict itself to actions whose outcomes are certain
C.Learn the effects of its actions through experience
D.Wait until the designer supplies a model before acting

Answer: C

Without a model the agent must explore, observe what its actions do and build a model as it goes, which is exactly the reinforcement learning setting. Waiting is not an option in an operating system, optimistic assumptions are unsafe, and confining action to certain outcomes is usually impossible. An environment that must be learned while it is acted in combines the exploration problem with the control problem, which is why it is the hardest of the standard cases.

Discuss in Forum

84.

Which of these is a percept for a taxi driving agent?

A.The decision to change lane at the next opportunity
B.The passenger's stated destination for the journey
C.The route selected by the navigation subsystem
D.A camera image of the road ahead

Answer: D

A percept is raw input from a sensor, and the camera image is exactly that. The lane change decision and the selected route are outputs of the agent's own reasoning, and the destination is arguably a percept when spoken but becomes a goal once adopted. Distinguishing what arrives from the environment from what the agent produces internally is the first discipline the framework imposes, and it is easy to blur in a system with many components.

Discuss in Forum

85.

Which of these is an actuator for a taxi driving agent?

A.The steering mechanism
B.The camera mounted on the windscreen
C.The map database stored in the vehicle
D.The route planning algorithm it executes

Answer: A

Actuators are whatever the agent uses to affect the world, so steering, throttle, brake, indicators and the display or voice used to speak to the passenger all qualify. Cameras are sensors, and the map and the planner are internal resources. Listing the sensors and actuators explicitly is part of the PEAS description, and doing so often reveals that the agent has less influence over its environment than the design assumed.

Discuss in Forum

86.

What might a poorly chosen performance measure for a taxi agent reward?

A.Choosing a route that avoids congested areas of the city
B.Speed at the expense of safety and passenger comfort
C.Arriving at the correct destination as requested
D.Obeying the traffic regulations that apply on the route

Answer: B

A measure that weights journey time heavily and safety lightly produces exactly the driving one would expect, which is why the measure must combine safety, legality, comfort, cost and speed with proportionate weights. The other three describe behaviours a good measure would encourage. Constructing a measure over several competing objectives is harder than it looks and is where most of the design difficulty in this example actually lies.

Discuss in Forum

87.

Why must a taxi agent's performance measure combine several objectives?

A.Multiple objectives make the measure easier to compute
B.Passengers cannot agree on which single objective matters
C.Optimising any one alone produces unacceptable behaviour
D.Regulations require every objective to be listed separately

Answer: C

Maximising speed alone gives dangerous driving, maximising safety alone gives a vehicle that never moves, and only a weighted combination captures what is actually wanted. Choosing those weights is a value judgement rather than a technical one, which is why it belongs to the designer and the operator rather than to the agent. Regulation, computability and passenger preference are secondary. Any objective that admits a degenerate optimum needs a competing term to balance it.

Discuss in Forum

88.

What is a stigmergic interaction between agents?

A.Direct message passing between agents in a common language
B.Coordination imposed by a central controller directing each agent
C.Agents observing one another's actions and imitating them
D.Coordination achieved by changing the shared environment

Answer: D

Ants coordinate by depositing pheromone that alters the environment other ants perceive, so complex collective behaviour emerges with no messages and no central control. Swarm robotics and ant colony optimisation borrow the idea directly. Messaging, central control and imitation are alternative coordination mechanisms. Using the environment itself as the communication channel is cheap, robust and scales to very large numbers, which is why it recurs wherever many simple agents must act together.

Discuss in Forum

89.

What is emergent behaviour in a multi-agent system?

A.Collective behaviour not explicitly programmed into any individual
B.Behaviour that appears only after the system has been running a long time
C.Behaviour that the designers did not intend and consider a fault
D.Behaviour produced by the most capable agent in the group

Answer: A

Flocking, foraging and traffic patterns arise from simple local rules interacting, and the collective pattern is nowhere written down in any agent's program. It may be desirable or not. Delay, undesirability and individual dominance are incidental. Emergence is attractive because it achieves complex outcomes cheaply and troubling because it is hard to predict or guarantee, and both aspects follow from the same absence of explicit specification.

Discuss in Forum

90.

Why is emergent behaviour difficult to guarantee?

A.Emergent behaviour changes randomly each time the system is run
B.The collective outcome is not stated anywhere to be verified against
C.Emergent systems always contain too many agents to simulate
D.Individual agents cannot be tested in isolation from the group

Answer: B

Since no component specifies the collective outcome, there is nothing to check a component against, and the behaviour must be established by simulation or by analysis of the dynamics. That is a genuine obstacle in safety critical applications. Scale, isolated testing and randomness are practical complications rather than the fundamental one. A property that exists only at the level of the whole cannot be verified by examining the parts, which is a general difficulty with emergent design.

Discuss in Forum

Loading…

Topics

Other Subjects

Tags: Intelligent Agents Questions and Answers || Intelligent Agents MCQ Questions and Answers || Intelligent Agents GK Questions and Answers || Intelligent Agents GK MCQ Questions || Intelligent Agents Multiple Choice Questions and Answers || Intelligent Agents GK || GK on Intelligent Agents || Artificial Intelligence Questions and Answers || Artificial Intelligence MCQ Questions and Answers || Artificial Intelligence GK Questions and Answers || GK on Artificial Intelligence

Intelligent Agents Questions and Answers

Topics

Other Subjects

About Us

Stay informed about all important Dates & Notifications