To move without bumping into a wall or a person, a robot needs to know where it is and what surrounds it. It builds a map of the environment and constantly compares what it sees with what it knows, updating its own position. It does this by combining several sensors, because none is perfect on its own: each technology has strengths and weaknesses.
Lidar
Lidar emits laser pulses and measures how long they take to come back after hitting an object. The result is a point cloud that accurately describes the distance and shape of obstacles. It is very reliable by day and by night, because it does not depend on ambient light. Its limits are cost, which makes it common mainly on more expensive robots, and trouble with some highly reflective or transparent surfaces, such as glass and mirrors.
Cameras and vision
Cameras recognise objects, surfaces and signs. With artificial intelligence they can tell a person from an object, or a rug from a bare floor, and work out what to do. Their weakness is that they depend on light: in the dark or against strong backlight, quality drops. That is why cameras are often paired with other sensors.
Proximity sensors
Ultrasonic, infrared or other short-range sensors detect nearby obstacles, the ones the robot is about to hit. They are cheap and reliable at short distances, and they handle last-moment decisions, such as stopping in front of a step or turning before a corner.
Inertial sensors and odometry
Accelerometers, gyroscopes and counting wheel rotations help the robot know how far it has moved and in which direction. On their own, however, they accumulate errors: a small inaccuracy adds up with every metre, and after a while the position is wrong. That is why they are checked against data from the other sensors.
Why the combination matters
Lidar provides the precise map, the camera recognises objects, short-range sensors avoid last-second collisions and odometry tracks movement. When one of these systems struggles, the others compensate. A robot that relies on a single technology will be more fragile in certain conditions, and that explains many of the situations in which a robot gets stuck or makes mistakes.
What to check when choosing
When you choose a robot, check which sensors it uses and in which conditions they work. A specification that only mentions cameras may hide limits in the dark, while a robot with lidar is more reliable in dim or unlit spaces. Also ask how it handles difficult surfaces such as mirrors, glass and glossy floors, because those are where performance varies most.
A robot’s perception is not magic: it is a set of sensors that correct one another. Understanding this helps you use it better and have realistic expectations about what it can do.