Home > Mobile Robotics Lab
Sensor Modelling

Why do Robots need Sensors?

As much as humans need sensors for making impressions of the world around, to navigate freely without colliding and sometimes even to localize ourselves with reference to the world around, robots need them for same reasons. Sensors enable robots to map the surroundings, to navigate without collisions and to localize themselves by recognizing presence of landmarks. Indeed sensing is a stepping stone to intelligence and autonomy and hence for any kind of autonomous or intelligent operations sensing and sensors are inevitable.

What kinds of Sensors are available?

Broadly sensors in robots are classified into two. They are

Proprioceptive sensor: Proprioceptive sensors perceive properties internal to the robot. Such sensors include battery capacity, wheel encoders, and position of arms and joints.

Exteroceptive sensors: Exteroceptive sensors perceive elements of the external world such as light levels, sound, and distance to objects. Hence Exteroceptive sensor measurements are interpreted by the robot in order to extract meaningful environmental features.

Some of the most common and popular sensors are:

a) Infrared Sensors (Proximity sensors)

b) Laser range finder (Range finders)

c) Sonar Sensors (Range finders)

d) Monocular Camera

e) Binocular Cameras (Depth finders)

Why do we need a Sensor Model?

Sensors by nature are noisy in the sense the measurements obtained from them cannot be taken at face value. For example when a range sensor measures the distance to an obstacle as say 1m it does not necessarily imply that the obstacle is 1m away from the sensor. A more likely interpretation of this measurement is that the obstacle is more likely to be around 1m from the sensor than say at 0.5m or 1.5m away. Or the probability that the obstacle is 1m away is more likely than the probability that it is 0.5m away.

Noises come due to various sources. In case of a monocular camera imaging a scene varying ambient illumination affects interpretation of the scene. If a pair of camera views are being used to detect depth then uncertainties in camera calibrations parameters such as its focal length can affect the depth being determined.

In case of a range sensor such as a laser range finder, the uncertainty in the range is known to increase as the distance to the object increases. In case of a sonar uncertainty arises due to its beam width or the divergence angle and also as distance to measured objects increase.