Skip to main content

Beginner Tier Exercises

Test Your Understanding of Perception, Sensors, and Navigation Concepts


Overview

These exercises test your understanding of the concepts from the Beginner tier:

  • B1: Introduction to Robotic Perception
  • B2: Understanding Sensor Types
  • B3: SLAM and Navigation Concepts

Complete these exercises before moving to the Intermediate tier.


Exercise 1: Perception Pipeline Ordering

Task: Put these perception pipeline stages in the correct order (1-4):

StageOrder (1-4)
Feature Extraction___
Interpretation___
Sensing___
Preprocessing___

Bonus: For each stage, give one example of what happens there.

Click for answer
  1. Sensing - Camera captures pixels, LIDAR measures distances
  2. Preprocessing - Noise filtering, calibration, synchronization
  3. Feature Extraction - Edge detection, object detection, clustering
  4. Interpretation - "This is a person," "This area is navigable"

Exercise 2: Sensor Selection Challenge

Scenario: You're building a robot for each situation below. Choose the most appropriate sensor(s) and justify your choice.

Scenario A: Delivery Robot in Office Building

  • Environment: Indoor, well-lit hallways
  • Tasks: Navigate to rooms, avoid people and furniture
  • Budget: Medium

Your answer:

  • Primary sensor: _______________
  • Secondary sensor: _______________
  • Justification: _______________

Scenario B: Agricultural Drone for Crop Monitoring

  • Environment: Outdoor, variable lighting
  • Tasks: Fly over fields, detect plant health by color
  • Budget: Low-Medium

Your answer:

  • Primary sensor: _______________
  • Justification: _______________

Scenario C: Robot Arm for Bin Picking

  • Environment: Indoor factory
  • Tasks: Identify and grasp objects of different colors/sizes
  • Budget: Medium-High

Your answer:

  • Primary sensor: _______________
  • Secondary sensor: _______________
  • Justification: _______________
Click for suggested answers

A: Delivery Robot

  • Primary: 2D LIDAR - reliable obstacle detection, good for SLAM
  • Secondary: RGB camera - for recognizing room signs/numbers
  • Why: Indoor environment suits LIDAR well; camera adds recognition capability

B: Agricultural Drone

  • Primary: RGB camera with multispectral option
  • Why: Need color for plant health analysis; must work outdoors where depth cameras struggle; lightweight for flight

C: Bin Picking Robot

  • Primary: RGBD camera (e.g., RealSense)
  • Secondary: Force/torque sensor at gripper
  • Why: Need both color (object identification) and depth (grasping location)

Exercise 3: SLAM Concept Questions

Answer these conceptual questions about SLAM:

Question 3.1

Why is SLAM called "the chicken-and-egg problem"?

Your answer: _______________

Question 3.2

A robot drives in a square and returns to start. Its odometry shows it's 0.5m away from where it started. What happened and how does SLAM handle this?

Your answer: _______________

Question 3.3

What does each color mean in a standard ROS 2 occupancy grid map?

  • White: _______________
  • Black: _______________
  • Gray: _______________
Click for answers

3.1: SLAM is chicken-and-egg because:

  • To build a map, you need to know where you are (to place observations correctly)
  • To know where you are, you need a map (to match observations against)
  • Neither can be solved first - they must be solved simultaneously

3.2: The 0.5m error is odometry drift from accumulated wheel slip and measurement errors. SLAM handles this through loop closure - when the robot recognizes it's back at the start, it adjusts the entire trajectory to eliminate the error.

3.3:

  • White = Free space (robot can pass)
  • Black = Occupied (wall/obstacle)
  • Gray = Unknown (not yet observed)

Exercise 4: Navigation Components Identification

Task: Match each Nav2 component to its correct description.

ComponentDescription
1. Map ServerA. Generates velocity commands to follow path
2. AMCLB. Computes global path from start to goal
3. Planner ServerC. Handles stuck situations (spin, backup)
4. Controller ServerD. Loads and publishes the map
5. Behavior ServerE. Estimates robot's pose on the map

Answers: 1=, 2=, 3=, 4=, 5=___

Click for answers

1=D, 2=E, 3=B, 4=A, 5=C


Exercise 5: True or False

Mark each statement as True (T) or False (F):

#StatementT/F
1RGB cameras provide depth information directly___
2LIDAR works well in complete darkness___
3Time-of-Flight cameras are best for outdoor use___
4Global planning happens once, local planning happens continuously___
5SLAM can only work with LIDAR sensors___
6Occupancy grids store the probability of each cell being occupied___
7The perception pipeline always ends with sensor data___
Click for answers
  1. False - RGB cameras only capture light intensity (color), not depth
  2. True - LIDAR uses its own laser light source
  3. False - ToF cameras struggle outdoors due to infrared interference from sunlight
  4. True - Global planner finds the path once (or when replanning), local planner runs continuously to follow it
  5. False - Visual SLAM uses cameras, RGB-D SLAM uses depth cameras
  6. True - Each cell stores 0 (free) to 1 (occupied), with 0.5 meaning unknown
  7. False - Pipeline ends with interpretation/understanding, not raw sensor data

Exercise 6: Diagram Interpretation

Look at the perception pipeline diagram from B1 and answer:

  1. What enters the pipeline at the "Sensing" stage?
  2. What exits the pipeline at the "Interpretation" stage?
  3. If a camera image has sensor noise, which stage removes it?
  4. If the robot needs to classify "is this a person?", which stage does that?
Click for answers
  1. Raw sensor signals from the physical world (photons, laser pulses, forces)
  2. Semantic understanding for decision-making (e.g., "obstacle at 2m", "person detected")
  3. Preprocessing - noise filtering happens here
  4. Interpretation - semantic classification happens here

Exercise 7: Practical Scenario Analysis

Scenario: A mobile robot is navigating through a crowded hallway. It has a 2D LIDAR and an RGB camera.

Answer these questions:

7.1 Perception

What can each sensor contribute to understanding the scene?

  • LIDAR: _______________
  • Camera: _______________

7.2 Navigation Challenge

A person suddenly walks in front of the robot's planned path. Trace what happens:

  1. Which sensor detects the person first?
  2. Which navigation component reacts?
  3. What does the robot do?

7.3 Failure Mode

The hallway has a glass door. What problems might occur and why?

Click for answers

7.1:

  • LIDAR: Distance to obstacles, wall positions, navigable space
  • Camera: Person recognition, door signs, color-based identification

7.2:

  1. LIDAR detects the person as a new obstacle (faster update rate)
  2. Local planner (Controller Server) detects obstacle in trajectory
  3. Robot either: stops, slows down, or finds a local detour around the person

7.3: Glass is problematic because:

  • LIDAR may pass through or give erratic readings (glass is semi-transparent to lasers)
  • Robot might plan a path through the door if it appears as free space
  • Solution: Use camera to detect visual features of the door, or add ultrasonic sensors

Self-Assessment Checklist

Before moving to the Intermediate tier, confirm you can:

  • Explain the four stages of the perception pipeline
  • Compare RGB cameras, depth cameras, and LIDAR
  • Describe what SLAM does and why it's challenging
  • Explain the difference between global and local planning
  • Identify the main components of Nav2
  • Understand occupancy grid maps

Next Steps

Congratulations on completing the Beginner tier exercises!

If you scored well on these exercises, you're ready for:

If you struggled with any topics, review the corresponding lesson before continuing.