The robot's senses

Four sensors cover most of what a small autonomous vehicle needs. Each has a character — strengths, blind spots, and a reason it exists.

Camera (with pan/tilt)

Rich, cheap, human-like — color, texture, signs, lines. But it is just pixels: distance must be inferred, darkness and glare hurt, and everything downstream (line following, detection) inherits its lighting sensitivity. The pan/tilt mount means where it looks is part of your design.

Lidar

Spins a laser and measures exact distances in a 360° plane — unaffected by lighting, perfect for obstacle detection and SLAM. Blind to color and texture, and it only sees its scan plane: a low object under the beam is invisible.

IMU (9-axis)

Accelerometer + gyroscope + magnetometer: orientation, acceleration, angular rate at high frequency. It feels tilts and turns instantly but drifts over time — great short-term, unreliable long-term.

Odometry

Integrates wheel motion into a position estimate (/odom). Smooth and always available, but errors accumulate — slips and drifts add up on long runs. Odometry answers "how far have I moved", not "where am I on the map"; anchoring to the map is SLAM's job. (The app's Reset also resets this estimate.)

The engineering habit

No sensor is trusted alone — real autonomy fuses them: camera for what, lidar for where, IMU for how it's moving, odometry for how far. When your robot behaves oddly, check what each sensor says (the app's panels) before blaming the code.