Module-4 Part-1

09-03-2026 JYO,AH,MEC 1
Navigation in Mobile Robots
• Navigation means A robot moving from one place to another safely
and intelligently to achieve a goal.
For example:
• A delivery robot moving inside a hospital
• A warehouse robot going to pick up a box
• A self-driving car reaching a destination
• Navigation is difficult because the robot must understand the world and
control itself correctly.
• Where am I? → Localization
• Where should I go? → Path Planning
• How do I move safely? → Obstacle Avoidance / Control
09-03-2026 JYO,AH,MEC 2
 The Four Building Blocks of Navigation
• Navigation is one of the most challenging competences required of a mobile robot.
• Success in navigation requires success at the four building blocks of navigation:
1. Perception (the robot must interpret its sensors to extract meaningful data).
2. Localization (the robot must determine its position in the environment).
3. Cognition (the robot must decide how to act to achieve its goals).
4. Motion control (the robot must modulate its motor outputs to achieve the desired
trajectory).

09-03-2026 JYO,AH,MEC 3
 Perception: understanding what is
happening
• The robot uses sensors like:
• Camera
• LiDAR
• Ultrasonic sensors
• GPS
• It must convert raw sensor data into meaningful information.
• Example:
• Detect walls
• Identify obstacles
• Recognize landmarks

Without perception, the robot is blind.

09-03-2026 JYO,AH,MEC 4
 Localization: Know where I am
• Localization means The robot must determine its position in the environment.
• The robot estimates:
• Its 𝑥 position
• Its 𝑦 position
• Its orientation 𝜃
Localization is difficult
• Because:
• Sensors are noisy( The measurements change slightly every time.)
• Wheels slip (instead of 1 meter, it moves 0.92m, due to slippery dusty floor)
• Measurements are imperfect (small errors in Angle measurements ,Encoder readings ,GPS
signals)
• That is why localization has received the most research attention in robotics.

09-03-2026 JYO,AH,MEC 5
• The most commonly used localization methods in modern robotics:

• Extended Kalman Filter (EKF)

• Particle Filter (Monte Carlo Localization)
• SLAM algorithms

• If the robot does not know its position:

• It cannot plan correctly

• It cannot reach its goal
• It cannot update its map

• So localization is the foundation of intelligent navigation.

09-03-2026 JYO,AH,MEC 6
 Cognition: Decide where to go
• This is the decision-making part.
• The robot must:
• Plan a path
• Avoid obstacles
• Choose actions to reach the goal
• Example:
• Should I turn left or right?
• Should I go around the obstacle?
• This is often handled using:
• Path planning algorithms
• Artificial Intelligence
• Reinforcement Learning
09-03-2026 JYO,AH,MEC 7
Motion control: Execute movement
• This converts decisions into motor commands.

• It ensures:
• Smooth movement
• Correct speed
• Accurate turning
• Trajectory tracking

• Even if planning is correct, bad motion control can make the robot crash.

09-03-2026 JYO,AH,MEC 8
09-03-2026 JYO,AH,MEC 9
Explanation of above figure 5.2:
Step 1: Start with an initial position
• The robot begins with an initial guess of its position inside the map.
• (The map database stores a structured representation of the environment
(walls, obstacles, landmarks, or occupancy probabilities) that the robot uses
to compare and correct its estimated position.)
Step 2: Move the robot
• The robot moves using its wheels.
Step 3: Measure wheel rotation (Encoder step)
• The encoders are sensors that measure how much each wheel rotated.
• From this, the robot calculates:
• How far it moved
• How much it turned
09-03-2026 JYO,AH,MEC 10
Step 4: Predict the new position
• Using the wheel movement, the robot estimates its new position.
• This is only a prediction, not exact.

Step 5: Sense the environment (Observation step)

• The robot uses sensors (like LiDAR or camera) to look at:
• Walls
• Corners
• Obstacles
• Landmarks

09-03-2026 JYO,AH,MEC 11
Step 6: Compare with the map (Matching step)
• The robot checks:If I am really at this predicted position, should I see
these walls and objects?
• It compares:
• Predicted position
• Sensor readings
• Map database
Step 7: Check the error
• The robot checks if the sensor readings match the map.
• If they match well → good prediction
• If they don’t match → position needs correction

09-03-2026 JYO,AH,MEC 12
Step 8: Correct the position (Update step)
• The robot adjusts its position estimate to reduce the error.
• Now it has a better estimate of (x, y, θ).

Step 9: Repeat forever

• The robot keeps repeating:
• Move → Predict → Sense → Compare → Correct → Repeat

09-03-2026 JYO,AH,MEC 13
 Reactive Navigation
• Reactive navigation is used in dynamic and unpredictable environments where
pre-planned paths may not be effective.
• It allows an agent (e.g., a robot) to adapt and react quickly to environmental
changes without relying on extensive prior knowledge or maps.
• Example: Obstacle Avoidance in Mobile Robots
• The robot uses sensors (such as ultrasonic, infrared, or LIDAR) to detect
obstacles.
• It employs reactive algorithms to make instantaneous decisions, such as:
• Adjusting speed
• Changing direction
• Performing evasive maneuvers

09-03-2026 JYO,AH,MEC 14
• Integration with Other Navigation Techniques: Reactive navigation can be combined with
deliberative planning (which involves strategic decision-making).
• It can also work with learning-based approaches, where the system learns from experience
to improve navigation.
• This combination provides a balance between immediate responses and long-term planning.
• Flexibility and adaptability: The agent interacts with a dynamic environment in real time.
• Common applications: Mobile robots, Autonomous vehicles, Other autonomous systems
that require quick decision-making.

09-03-2026 JYO,AH,MEC 15
Localization and challenges of localization-
Assignment 2 topic
• Please study from the text book I have uploaded in googleclassroom.
Module 3 &4 – same text book.

09-03-2026 JYO,AH,MEC 16
An error model for odometric position
estimation

09-03-2026 JYO,AH,MEC 17
An error model for odometric position
estimation
• Odometry means Estimate where the robot is by measuring how
much the wheels have moved.

09-03-2026 JYO,AH,MEC 18
• For a differential-drive robot ,If we know:
• Starting position

• How much the robot moves in small time intervals

• Then we can find the new position by:

• Adding up small movements step by step

• This is called integration of motion

(or summing incremental distances).

09-03-2026 JYO,AH,MEC 19
09-03-2026 JYO,AH,MEC 20
09-03-2026 JYO,AH,MEC 21
• Δ𝑠𝑟 ∶Distance traveled by the right wheel during the small time interval
Δ𝑡
• Δ𝑠𝑙 : Distance traveled by the left wheel during the same time interval

09-03-2026 JYO,AH,MEC 22
09-03-2026 JYO,AH,MEC 23
09-03-2026 JYO,AH,MEC 24
09-03-2026 JYO,AH,MEC 25
• The error propagation model for odometry.

09-03-2026 JYO,AH,MEC 26
09-03-2026 JYO,AH,MEC 27
09-03-2026 JYO,AH,MEC 28
09-03-2026 JYO,AH,MEC 29
Note:

09-03-2026 JYO,AH,MEC 30
09-03-2026 JYO,AH,MEC 31
In matrix ,

• Right and left wheel errors are assumed independent.

• So no correlation between them.
• If one slips, we assume it does not affect the other.

09-03-2026 JYO,AH,MEC 32
Variance

09-03-2026 JYO,AH,MEC 33
After finding uncertainty/covariance, then we propagate error to

09-03-2026 JYO,AH,MEC 34
09-03-2026 JYO,AH,MEC 35
• summary

09-03-2026 JYO,AH,MEC 36
09-03-2026 JYO,AH,MEC 37
Jacobian with Respect to Pose 𝐹𝑝 ((Eq. 5.10))

09-03-2026 JYO,AH,MEC 38
Jacobian with Respect to Wheel Motion
𝐹Δ ((Eq. 5.11)

09-03-2026 JYO,AH,MEC 39
09-03-2026 JYO,AH,MEC 40
09-03-2026 JYO,AH,MEC 41
09-03-2026 JYO,AH,MEC 42