Robotics Perception


What makes the difference between robots and simple mechanisms is the ability of robots to adapt to changes of their subjects of operation or of their operating environment. This entails the robot ability
  1. to understand the surrounding environment,
  2. to derive a set of actions from the high level goal the robot has been given and
  3. to implement (actuate and control) these actions.
 Robotics perception is used both in 1. and 3. The perception is phisically implemented by sensors and by dedicated processing of the data they produce.
Sensors can be classified in
  • enteroreceptive sensors, that sense things inside the robot (e.g. joint angle,speed, torque) and
  • exteroreceptive sensors, that sense things outside the robot (e.g. proximity, vision).

  •  
     
     
     
    The DEXARM robot joint
     
     
    Enteroreceptive Sensing
     
    Enteroreceptive sensors are used by the robot control systems to receive feedback on the execution of motion and in general on the status of the robot. For what concerns these sensors, the R&D made by the A&R Section has mainly focussed on the tight integration of the sensor with the actuation means in a compact robotics joint.
     
     
     
    Breadboard of Imaging System for BepiColombo MSE
     
     
    Exteroreceptive Sensing
     
    Exteroreceptive sensors provide the robot control system information:
    • about the environment around the robot (e.g. rover Cameras provide images of the terrain around the rover) and
    • about the effect of robot actions on the environment (e.g. the distance between a robot hand and the object it grasps)
    The information provided by the sensors needs to be processed for the robot to understand it. For example images from rover cameras arranged in stereo pairs are prodessed to create 3D maps where a rover control can identify dangerous obstacles. The A&R section has focussed its R&D efforts on the processing part, mainly in the field of Computer Vision.
     
     
     
    Last update: 6 July 2007


    More in depth ...

     •  Vision manipulation of non-cooperative objects (http://www.esa.int/TEC/Robotics/SEMHHQNSP3F_0.html)