Proprioceptive Sensors

These sensors monitor the robot's internal state and position.

Encoders:

• Measure joint angles and positions
• Rotary: Track rotation amount and direction
• Linear: Measure linear displacement
• Resolution: 0.01° to sub-micrometer level
• Applications: All robotic arms, wheeled robots

Inertial Measurement Units (IMU):

• Accelerometers: Measure acceleration (3-axis typically)
• Gyroscopes: Measure rotational velocity
• Magnetometers: Measure magnetic field (compass)
• Combined IMU: 9-axis systems (3 accel + 3 gyro + 3 mag)
• Applications: Humanoid robots, drones, mobile robots

Force/Torque Sensors:

• Measure forces and moments
• 6-axis systems: All forces and moments
• Precision: 0.01N to high accuracy
• Applications: Manipulation, assembly, haptic feedback

Other Proprioceptive:

• Motor current sensors (torque indication)
• Temperature sensors
• Pressure sensors
• Vibration sensors

Exteroceptive Sensors

These sensors measure the external environment.

Vision Systems:

• RGB Cameras: Standard color imaging
• Depth Cameras: Measure distance to objects (ToF, structured light, stereo)
• Thermal Cameras: Infrared imaging
• Hyperspectral: Multiple wavelengths for material identification
• Industrial Cameras: 30 MP to 8K resolution

Range Finders:

• LiDAR: Laser scanning for 3D mapping
  • 2D: Single plane scanning
  • 3D: Multi-layer point clouds (16, 32, 64+ channels)
  • Range: 10-200+ meters
  • Resolution: 1-10cm at distance
• Ultrasonic: Sound-based distance measurement
  • Range: 2cm to 4+ meters
  • Cost: Very inexpensive
  • Applications: Proximity detection
• Infrared Proximity: LED-based distance
  • Short range (30cm typical)
  • Fast response
  • Good for collision avoidance

Tactile Sensors:

• Pressure arrays: Detect contact patterns
• Force-sensing skin: Distributed pressure
• Slip sensors: Detect object movement
• Temperature: Contact temperature measurement
• Applications: Grasping, manipulation, assembly

Audio Sensors:

• Microphone arrays: Directional sound
• Acoustic detection: Sound event recognition
• Speech recognition: Voice commands
• Applications: Human interaction, audio-based detection

Active vs Passive Sensing

Active Sensors:

• Emit energy into the environment
• Measure response to detect objects
• Examples:
  • LiDAR (laser light)
  • Structured light (pattern projection)
  • Sonar (sound waves)
  • Radar (radio waves)
• Advantages: Works in darkness, well-defined measurement
• Disadvantages: Can interfere with other sensors, requires power

Passive Sensors:

• Detect ambient energy (light, heat, sound)
• No emission required
• Examples:
  • RGB cameras
  • Thermal cameras
  • Microphones
• Advantages: Low power, no interference, natural
• Disadvantages: Lighting dependent, less precise range

Path & Motion Planning

Path Planning:

• Find collision-free route from start to goal
• Algorithms:
  • A* Search: Optimal for grid worlds
  • Rapidly-exploring Random Trees (RRT): For high-dimensional spaces
  • Dijkstra's Algorithm: Shortest path
  • Potential Field: Gradient-based approach
  • Probabilistic Roadmap (PRM): Pre-computed graph

Motion Planning:

• Generate smooth trajectories
• Consider dynamics and constraints
• Coordinate multiple joints/actuators
• Time optimization
• Methods:
  • Trajectory interpolation
  • Optimal control theory
  • Spline-based curves
  • Minimum energy paths

Example: A 6-axis robot reaching from point A to point B:

1. Inverse kinematics: Calculate joint angles
2. Collision checking: Verify no obstacles
3. Trajectory planning: Smooth joint-space path
4. Time parameterization: Add timing information
5. Execution: Send commands to motors

Control Systems

Feedback Control:

• Monitor actual performance
• Compare to desired performance
• Adjust actions based on error
• Classic: PID (Proportional-Integral-Derivative)

Control Loop Example:

Goal: Position arm at X coordinate = 50 cm
Current: X = 45 cm (error = 5 cm)
Action: Move motor forward
Measure: New X = 49 cm (error = 1 cm)
Action: Fine-tune motor speed
Until: Error approaches zero

Types of Control:

• Position Control: Reach specific location
• Velocity Control: Maintain speed
• Force Control: Apply specific force
• Impedance Control: Spring-like behavior
• Hybrid Control: Switch between modes

Advanced Control:

• Model Predictive Control (MPC): Look ahead
• Adaptive Control: Adjust to changing conditions
• Robust Control: Handle uncertainty
• Optimal Control: Minimize cost function

Machine Learning & Adaptation

Supervised Learning:

• Train from examples with labels
• Learn from human demonstrations
• Image classification: Object identification
• Speech recognition: Voice commands

Reinforcement Learning:

• Learn by trial and error
• Reward for good actions
• Policy optimization
• AlphaGo, robot manipulation learning

Unsupervised Learning:

• Find patterns in data
• Clustering similar situations
• Feature extraction
• Anomaly detection

Transfer Learning:

• Use knowledge from one task for another
• Pre-trained models (ImageNet, GPT)
• Domain adaptation
• Sample-efficient learning

Deep Learning:

• Neural networks for complex patterns
• CNNs for vision
• RNNs for sequences
• Transformers for large-scale learning

Reasoning & Knowledge

Symbolic Reasoning:

• Logic and rules
• If-then statements
• Knowledge graphs
• Ontologies

Common Reasoning Types:

• Deductive: General rule → specific case
• Inductive: Specific cases → general rule
• Abductive: Observations → most likely explanation
• Analogical: Similar situation → apply known solution

Uncertainty Handling:

• Probabilistic reasoning
• Bayesian networks
• Markov decision processes
• Fuzzy logic

Example: Robot reasoning for object grasping:

1. Observe: Red, cylindrical object
2. Recall: Cylinders best grasped with parallel gripper
3. Plan: Approach from top with parallel gripper
4. Execute: Lower gripper and grasp
5. Learn: Save successful grasp pattern