AS-2425-lecture3_sensors

Autonomous Systems Lecture 3 – Simple and Complex Sensors

Introduction

• Material based on work by:

  • Dr. Marco Wiering

  • Dr. Matias Valdenegro

  • Henry Maathuis

  • Jelle Visser

  • Ben Wolf

Previous Lecture Summary

• Students should understand basics of:

  • Components of a robot

    • Sensors

    • Effectors and actuators

    • Controllers

  • Degrees of Freedom (DOF)

  • Robot locomotion:

    • Legged locomotion

    • Gaits

    • Wheeled locomotion

  • Trajectory planning

Today's Lecture Goals

• Outline sensors commonly used in robotics.

• Understand differences between simple and complex sensors.

• Understand how sensors are used for perception.

• Introduce concepts from machine vision.

• Show examples:

  • Face recognition

  • Emotion recognition

  • Gesture recognition

Sensor Overview

Simple and Complex Sensors

• Simple Sensors: Provide basic data (1D) without requiring further processing.

• Complex Sensors: Provide multidimensional data, necessitating complex processing for utility.

Robot Perception

• Robots perceive:

  • Environment (exteroception)

  • Other agents and actions

  • Themselves (proprioception)

• Importance of perception:

  • Knowing the state

  • Knowing possible actions

  • Reward estimation

  • Gauging priorities

Sensor Terminology

• Active vs Passive Sensors:

  • Active Sensors: Emit a signal, interact with the environment for measurement. e.g. sonar or ultrasound. Active sensors emit energy

  • Passive Sensors: Measure without direct interaction (e.g., light sensors).

• Simple vs Complex Sensors:

  • Simple: Basic functions (e.g., switches, light sensors).

  • Complex: Advanced functions needing processing (e.g., ultrasound, cameras).

Examples of Simple Sensors

Switches

• Binary output, detects collisions, and motion limits.

• Simples sensor

Light-sensitive Diode

• Produces a signal from light exposure via voltage/resistance.

• Used in daylight sensing and optical media reading.

Position Sensors

• Detect linear/angular positions/velocity using potentiometers or pattern observation.

• Applications include odometry (wheel position/rotation) and joint position determination.

• How do you get rotation from position?

  • by adding rings with different codes > bit encoding

  • adding more bits > higher precision

Servo Mechanism

• Measures and controls rotational position using:

  • Rotary encoder sensor

  • Gearbox

  • Motor as an actuator

• Applications in robotics such as actuating small components.

Accelerometers

• Measures acceleration using Newton's law (F = ma).

• Used for velocity measurements, orientation, and position estimation.

Gyroscopes

• Measures angular position through conservation of angular momentum.

• Typically combined with accelerometers for better orientation estimates.

Complex Sensor Examples

Ultrasonic and Sonar Sensors

• Use sound frequencies for echolocation, inspired by bats and dolphins.

• They require processing for meaningful information.

• Can be non-precise as the waves bounce of other surfaces thus generating false readings.

Laser Sensors (LiDAR)

• Use light beams for measuring reflections, we use phase-shift to measure information.

• Generate point clouds with unidirectional laser beams.

Vision in Robots

• Camera Usage: Mimics human vision but with differences:

  • Resolution non-uniformity.

  • Processing variance between luminance and color data.

• General categories include:

  • Early vision: Basic image representation.

  • High-level vision: Further processed data.

Depth Sensing

• Devices like Kinect and Intel RealSense produce RGBD images ( different colours depending on the depth)

• Help estimate distances in robotic applications.

Machine Vision Use Cases

• Object detection and classification.

• Face and emotion recognition aiding human-robot interaction.

• Motion vision: Differentiate frame changes for focus.

Neural Network- based vision

• nowadays using deep neural networks

  • Classification – what is it?

  • Object detection – is there something?

  • Localisation – where is something?

  • Segmentation -- what belongs to something?

Feature Extraction and Interpretation

• Simplifying vision involves filtering information from images.

• Utilizes color tracking and motion detection combined.

• Questions for selecting features:

  • Task specificity

  • Environment distinctiveness

  • Sensor availability

  • Computational intensity

Summary

• Sensors can be simple/complex, active/passive based on applications.

• Vision and sensor fusion necessitate complex processing to derive meaningful features from data.