RobotGuide
Hardware selection/Sensor

Selection 3 / 5

Sensor

What to measure, how fast, how accurate

Choosing a sensor only takes three questions: what physical quantity, how fast, how accurate. Once those three are fixed, the model is basically unique. Picking a model first and thinking about the use later is the most common beginner waste.

SensorMeasuresRateAccuracy / resolutionPriceLimits
Incremental encoderRelative position20 kHz+Depends on line count (1k–10k lines)¥50–300Must home on power-up
Absolute encoderAbsolute position10 kHz17–23 bit¥300–2000Expensive; protocol varies by vendor
MEMS IMUAngular rate + acceleration200–1000 HzLarge drift (°/h level)¥30–300Integration drifts; fusion is mandatory
Industrial / fiber-optic IMUAttitude200 HzSmall drift¥3000+Expensive, heavy
Torque sensorJoint torque1–5 kHz0.1% FS¥1000+Needs calibration; temperature drift
Single-line lidar2D distance10 Hz±2 cm¥300–1000No height information
Depth cameraColor + depth30 Hz±1–2 cm @1 m¥1000–4000Fails in strong light / reflections / transparent objects
StereoPassive ranging30 HzDegrades with distance squared¥500–3000Fails in weak-texture scenes
Six-axis force/torqueForce + torque1 kHz0.5% FS¥3000+Expensive; demands a stiff mounting surface

Quick picks for three scenarios

You need a 1 kHz current loop

Encoder + phase-current sampling

Do not expect an IMU or vision to join the inner loop

You need grasping

Depth camera + end-effector force/tactile

Vision gives pose; force tells you whether it actually gripped

You need navigation

Single-line lidar + IMU + wheel odometry

Three complementary channels: lidar gives geometry, IMU gives attitude, wheel speed gives short-term displacement

Three easy-to-miss points