Unit 6: AI for Intelligent Systems and Robotics - Practice Quiz

1 What is the main purpose of image classification?

2 What does object detection identify in an image?

3 Which task is an example of image classification?

4 What is a bounding box used for in object detection?

5 What does Natural Language Processing mainly help computers understand?

6 Which is an example of an NLP application?

7 What is a chatbot?

8 What is the purpose of conversational AI?

9 What is text classification in NLP?

10 In reinforcement learning, what is an agent?

11 What does the state represent in reinforcement learning?

12 What is an action in reinforcement learning?

13 What does a reward provide to an agent?

14 A robot receives a positive reward for reaching a target. What should it learn to do?

15 What is an autonomous robot?

16 Which capability helps an autonomous robot avoid obstacles?

17 Why do autonomous robots use computer vision?

18 Which task can AI help an autonomous robot perform?

19 Which is a likely future trend in intelligent robotics?

20 What does explainable AI aim to provide?

21 A warehouse robot receives an image containing three boxes and two workers. It must identify each object's category and location. Which computer vision task is most suitable?

22 An inspection robot classifies product images correctly during training but performs poorly on images captured under different factory lighting. Which change would most directly improve robustness?

23 A robot must detect small tools scattered across a workbench. The detector frequently misses tools that occupy very few pixels. Which adjustment is most likely to help?

24 An object detector predicts a box for a chair. The predicted box overlaps the labeled box, but their overlap is small compared with their combined area. What does this imply about the Intersection over Union (IoU)?

25 A service robot receives the command, "Take the package to the reception desk." Which NLP output is most useful for planning the task?

26 A robot hears, "Place it beside the blue container," but two recently mentioned objects could be referred to by "it." What should the conversational system do?

27 A multilingual guide robot must understand visitor requests in several languages while using one intent classifier. Which approach is most appropriate?

28 A chatbot controlling a home robot correctly identifies the intent "set temperature" but cannot execute "Set it to 22 degrees." Which missing capability is the most likely cause?

29 A cleaning robot uses reinforcement learning. At a particular moment, its sensor readings indicate its position, battery level, and nearby obstacles. Together, these values represent the robot's:

30 A delivery robot earns for completing a delivery, for a collision, and for every minute used. What behavior does this reward design encourage?

31 A robotic arm always selects the action with the highest current estimated value and never tries alternatives. What learning problem may result?

32 A robot receives a reward of now and a reward of one step later. If the discount factor is , what is the two-step discounted return?

33 An autonomous mobile robot detects a person unexpectedly entering its planned path. Which response best demonstrates closed-loop autonomy?

34 A robot uses wheel odometry for navigation, but its estimated position gradually becomes inaccurate because the wheels sometimes slip. Which additional capability best addresses this problem?

35 A farm robot must navigate through a field it has never seen while building a map and estimating its own location. Which technique is most relevant?

36 An autonomous drone's navigation model works well in simulation but becomes unstable in real wind. Which training strategy most directly reduces this simulation-to-reality gap?

37 A next-generation household robot combines camera input, spoken instructions, and touch sensor readings to decide how to handle an object. Which emerging trend does this illustrate?

38 A medical-assistance robot processes camera data on the device instead of sending every frame to a cloud server. What is the main advantage of this edge AI approach?

39 A collaborative robot learns a new assembly task from a few worker demonstrations and then adapts when the product shape changes. Which future capability is most clearly represented?

40 Before deploying an AI robot in a public hospital, engineers test whether its decisions differ unfairly across patient groups and record why each action was selected. Which future priority does this address?

41 A warehouse robot must distinguish between visually similar packages while also locating several packages in the same camera frame. Which model output is most appropriate for this requirement?

42 An object detector performs well on daytime validation images but fails when a mobile robot enters a dim warehouse. Training and validation images have similar object categories, but their illumination distributions differ substantially. Which intervention most directly addresses the problem?

43 A safety-critical robot uses an object detector to identify humans. Missing a person is substantially more dangerous than producing an occasional false alarm. Which evaluation strategy is most appropriate?

44 Two predicted bounding boxes overlap the same ground-truth object with intersection-over-union values of and . A detector uses non-maximum suppression with an IoU threshold of . What normally happens?

45 A robot's classifier achieves accuracy on a dataset where only of images contain a dangerous obstacle. It correctly detects of dangerous images but incorrectly flags of safe images. Approximately what is the precision of the danger detector?

46 A service robot receives the utterance, "It is cold here; could you close that?" The word "that" refers to a nearby window identified by the robot's vision system. Which NLP capability is most essential for correct interpretation?

47 A chatbot receives the phrase "I need to cancel my appointment tomorrow," but the user has two appointments scheduled for tomorrow. What is the safest conversational response?

48 A robot assistant is deployed in a hospital. Its language model is highly fluent but sometimes invents medication instructions when retrieval fails. Which design change most directly reduces this safety risk?

49 A multilingual robot receives commands in languages not represented during training. Which approach is most likely to improve intent recognition while preserving a shared downstream policy?

50 A voice-controlled robot must distinguish between "pick up the red cup" and "do not pick up the red cup." Which failure would be especially dangerous if the NLP pipeline relies mainly on keyword matching?

51 A robot navigates using a camera that cannot observe obstacles behind walls. If the controller treats the current image as the complete state, which assumption is most likely violated?

52 A delivery robot receives a large positive reward when it reaches its destination but no intermediate rewards. Training is slow because successful trajectories are rare. Which modification most directly addresses this issue without changing the final objective?

53 A robot receives reward for reaching a target and for every time step. It can choose a risky shortcut or a safe longer route. What behavior does the reward function explicitly encourage?

54 During training, a robot learns to repeatedly approach a charging station and reset because the reset transition accidentally provides a large reward. What is the primary problem?

55 A robot's policy performs well in simulation but collides with furniture on a real floor because the simulated wheel friction and sensor noise were too idealized. Which strategy most directly improves transfer?

56 An autonomous delivery robot detects a pedestrian using vision, estimates the pedestrian's motion, plans a path, and sends velocity commands. Which architectural property is most important for safe operation under changing conditions?

57 A robot has an accurate global map but its localization estimate drifts when wheel slip occurs. Which sensor-fusion improvement is most appropriate?

58 A robot must operate near humans, but its perception model is uncertain because of rain and glare. Which decision policy is most appropriate for a safety-critical action?

59 A future household robot is expected to perform novel tasks described in natural language and adapt to unfamiliar homes. Which combination of capabilities most directly supports this goal?

60 A robot manufacturer wants to deploy an AI model on an embedded device with limited power and strict response-time requirements. Which trend most directly addresses this constraint?