Interoceptive Sensing: Perceiving Internal State
Interoceptive sensing (also called proprioceptive sensing) refers to a robot's ability to measure its own internal state — the configuration, motion, and health of its own body. Just as humans feel the position of their limbs without looking at them, robots use interoceptive sensors to track their own joint angles, velocities, motor currents, and internal temperatures.
Common interoceptive sensors include:
- Joint encoders — measure absolute or incremental angular position of each joint
- Inertial Measurement Units (IMUs) — measure linear acceleration and angular velocity of the robot body
- Current sensors — measure motor drive current, which correlates to joint torque
- Temperature sensors — monitor motor and electronics thermal state
- Joint torque sensors — directly measure reaction torques at joints (used in collaborative robots)
Interoceptive data is typically available at high rates (1 kHz or more) with low latency, making it ideal for fast, closed-loop joint-level control. Because these sensors measure the robot's own body, they are not affected by occlusion, lighting conditions, or environmental clutter.
Exteroceptive Sensing: Perceiving the External World
Exteroceptive sensing refers to sensing that gathers information about the environment external to the robot. These sensors detect the world around the robot — obstacles, objects, people, terrain, light, and sound — enabling the robot to navigate, manipulate, and interact safely.
Common exteroceptive sensors include:
- Cameras — capture visual information (monocular, stereo, RGB-D)
- LIDAR — measures distances to surfaces using laser pulses, producing 2D or 3D point clouds
- Ultrasonic sensors — emit sound pulses and measure time-of-flight for proximity detection
- Tactile sensors — detect contact forces and pressure distribution at the robot's surface
- GPS / RTK-GPS — provide global position estimates for outdoor mobile robots
- Microphones / acoustic sensors — detect sound events or measure acoustic properties
Exteroceptive sensors are generally more susceptible to environmental conditions — cameras fail in darkness, ultrasonic sensors are confused by soft surfaces, and GPS is unavailable indoors. This motivates the use of multiple complementary sensor modalities.
Interoceptive vs. Exteroceptive Sensors
Interoceptive (Proprioceptive)
- Measures internal robot state (joints, IMU, currents)
- Examples: rotary encoders, IMUs, current sensors, joint F/T sensors
- High update rates (1 kHz+), low latency
- Not affected by environmental lighting or occlusion
- Used in low-level joint and balance control loops
- Errors accumulate over time (integration drift, backlash)
- Essential for precise trajectory execution
Exteroceptive
- Measures external environment (objects, terrain, people)
- Examples: cameras, LIDAR, ultrasonic, tactile, GPS
- Lower update rates (10–100 Hz), higher latency
- Susceptible to environment conditions (lighting, weather, surface material)
- Used in high-level perception, navigation, and manipulation planning
- Provides absolute reference — corrects accumulated drift
- Essential for obstacle avoidance, object recognition, localization
Sensor Classification Reference Table
| Sensor Type | Category | Measured Quantity | Typical Use |
|---|---|---|---|
| Rotary encoder | Interoceptive | Joint angle / angular velocity | Joint position control |
| IMU (gyro + accel) | Interoceptive | Angular velocity, linear acceleration | Balance, odometry, SLAM |
| Motor current sensor | Interoceptive | Drive current (proxy for torque) | Torque control, collision detection |
| Joint torque sensor | Interoceptive | Reaction torque at joint | Force control, impedance control |
| Temperature sensor | Interoceptive | Motor / electronics temperature | Thermal protection |
| Monocular camera | Exteroceptive | RGB image (scene appearance) | Object detection, visual servoing |
| Stereo camera | Exteroceptive | RGB image + depth map | Navigation, 3D reconstruction |
| RGB-D camera | Exteroceptive | Color image + depth (structured light / ToF) | Manipulation, SLAM |
| 2D LIDAR | Exteroceptive | Range scan (single plane) | Mobile robot navigation, mapping |
| 3D LIDAR | Exteroceptive | Dense 3D point cloud | Autonomous vehicles, inspection |
| Ultrasonic sensor | Exteroceptive | Distance (time-of-flight of sound) | Proximity detection, collision avoidance |
| GPS / RTK-GPS | Exteroceptive | Global position (lat/lon/alt) | Outdoor localization |
| Tactile array | Exteroceptive | Pressure distribution at contact | Grasping, manipulation |
Key Principles of Sensor Classification
- 01
Robots require both interoceptive and exteroceptive sensing to operate effectively — internal state alone is insufficient for navigation, and external perception alone cannot drive joint-level control.
- 02
Active sensors emit energy (laser pulses, ultrasonic waves, structured light) and detect the reflected return. Passive sensors (cameras, thermometers) only detect naturally occurring energy.
- 03
Active sensors are generally more robust in varying lighting but consume more power and can interfere with other active sensors of the same type.
- 04
Sensor fusion between interoceptive and exteroceptive modalities corrects accumulated drift: e.g., wheel odometry (interoceptive) drifts, but LIDAR loop closure (exteroceptive) resets the error.
- 05
Dead reckoning uses only interoceptive sensors (wheel encoders, IMU) to estimate pose — errors grow unbounded without exteroceptive corrections.
- 06
Closed-loop control requires sensor feedback; the choice of sensor category determines the control bandwidth achievable.
- 07
Sensor placement is a design decision — interoceptive sensors are embedded in actuators; exteroceptive sensors are placed for maximum field of view and minimum occlusion.