Exteroceptive and Proprioceptive Sensors for Robotics

Proprioception refers to the human perception, conscious or not, of the position of different parts of the body. Exteroception, as opposed to proprioception, includes sensations caused by external stimuli (sight, smell, hearing…). It is these phenomena that allow man to move around in an environment and to be aware of its location.

What are the different types of sensors?

According to this same principle, mobile robots (i.e. capable of locomotion, as opposed to manipulator robots) embody different types of proprioceptive and exteroceptive sensors to perform different functions.

In order to differentiate between these two categories of sensors, it is first necessary to define the general characteristics that are important for their use in mobile robotics.

All sensors are characterized by an acquisition frequency (response time), a measurement resolution, a noise on the physical quantity measured (defined by its repeatability) … Also, the measuring principle defines whether the sensor performs an absolute measurement (always in relation to a known reference frame) or relative.

The above characteristics are generally associated with the intrinsic measuring principle of each sensor and/or the acceptable cost of these sensors. For example, there are different technologies of distance measuring sensors (laser, ultrasonic, etc.) with different properties (resolution, reflectivity…) and a different cost.

Us sensor vs. TOF sensor. (Sensors not representative of the material used at Innowtech).

Proprioceptive and Exteroceptive Sensors

Proprioceptive sensors are used to measure the state of the robot itself (wheel position or speed, battery charge, etc.) while exteroceptive sensors are used to measure the state of the environment (cartography, temperature, etc.).

From the roboticist’s point of view, exteroceptive sensors are mainly absolute measurement sensors, with a generally lower acquisition frequency than proprioceptive sensors. The best known example is probably GPS.

However, this rule is not a general truth but may also depend on the use of the sensor.

Exteroceptive and proprioceptive sensors in practice

Let’s take the case of the accelerometer (measurement of an acceleration); the latter can be defined as exteroceptive if it measures the acceleration due to gravity, while it will be defined as proprioceptive if it measures the acceleration of a robot.

The difference between absolute and relative measurements is fundamental for a robot. Absolute measurements have a fixed and bounded measurement uncertainty (often greater than that of relative sensors) while relative measurements (often with low measurement uncertainties) are used cumulatively with odometry. As a result, relative sensors sooner or later end up providing information that is increasingly derived and unbounded.

One solution to this problem is to combine these different types of sensors (this is called sensor fusion) in order to obtain data that is more reliable (with a bounded uncertainty) and as accurate as possible. This objective can be illustrated by the use of inertial measurement units (or IMU).

Example of IMU

This object generally includes accelerometers, gyrometers (relative, proprioceptive, measuring angular velocity) and a magnetometer (absolute, exteroceptive, measuring angle relative to magnetic north). The magnetometer (also called magnetic compass) compensates the gyrometer drift (along the yaw axis) but is not able to detect angular movements as finely as the gyrometer. Accelerometers compensate the gyrometer drift along the pitch and roll axes by deducing the direction of the Earth’s gravity (absolute measurement). In this case, the fusion of proprioceptive and exteroceptive sensors allows to obtain very frequent, stable and precise orientation data.

Illustration du Lacet/Tangage/Roulis

Exteroceptive and proprioceptive sensors at INNOWTECH

Our robot-sensors evolve in constraining environments thanks to a judicious fusion of proprioceptive and exteroceptive sensors by combining the advantages of each one and making it possible to meet the constraints of severe environments and metrological needs (related to measurement accuracies).

Given the constraints of harsh environments, some sensors are prohibited (e.g.: impossibility to use a GPS inside a nuclear power plant). It is therefore necessary to study each case to define the appropriate combination of sensors in terms of desired accuracy, use in a constrained environment, and cost.