SPARCO


   
The multi-sensorized DLR End Effector
 
The goal of the SPAce Robot COntroller (SPARCO) project has been to develop a space robot controller based on a reliable, widespread, state-of-art robot controller extensively used in industry (automotive applications): the COMAU C3G-9000.

This industrial controller has been upgraded with Advanced sensor-based robot capabilities, as identified during previous ESA studies. The DLR End Effector, which is a highly sensorized smart end effector, already tested in space during the D2 mission, has been interfaced to the controller to provide additional sensing capabilities (Force / Torque, distance, tactile).

This approach has been preferred to the development of an "ad-hoc" space robot controller for two main reasons:
  • The use of a solid industrial platform allows the exploitation of already present robot control capabilities, thus avoiding re-developments and guaranteeing the high reliability which comes from extensive testing and day-by-day use of the system on a large scale in different applications.
  • The introduction of advanced control concepts and the use of smart sensors make the system compliant with most demanding performance requirements set by space applications.
The activity has been carried out, under ESA TRP contract by Tecnospazio (IT, Prime contractor), IPK Berlin, DLR-Oberpfaffenhofen (D), Fokker Space and Systems (NL).
 
 
 
Main result of this activity is the production of a SPARCO breadboard and its integration in the CAT testbed at ESTEC Robotics Laboratory.

For what concerns sensor-based control techniques, impedance control was considered; this technique allows the execution of most in-contact tasks with relatively low implementation effort.

The system was developed to accommodate, in the future, also other sensor-based control techniques, as hybrid force-position control and control based on distance and proximity sensors. Furthermore, to easy program development, the user of SPARCO breadboard is given a set of library, to perform basic in contact operations.
 
 
   
CAT Testbed
 
Activity Analysis
 
According to ESA top-level requirements for the designated application / scenario (CAT Testbed), the system shall be capable to execute a set of in-contact tasks like ACTUATE, OPEN, CLOSE. INSTALL, REMOVE, with defined performances.

Aim of Activity Analysis is to break down such tasks in Actions. An Action is uniquely mapped to a well precise control concept. There are two many action categories:
  • robot free motion: mapped to pure position control (e.g. DISPLACE TO Action);
  • robot motion in contact with environment: Force / Torque control (e.g. INSERT Action).
The modularity of this approach allows the user to extend, according to the needs of the new application, the task and action library. For example the OPEN task, applied to a sliding drawer, has been decomposed in the following ACTION sequence: DISPLACE, APPROACH, ATTACH, SLIDE, DETACH, RETRACT, DISPLACE.

After having characterized actions with specific attributes, each action was further decomposed in smaller parts, corresponding to already existing or new functionalities to be added in the robot control Software.
 
 
 
SPARCO Impedance Control Scheme
 
 
Control concept definition
 
The major effort in the SPARCO activity was to implement in the already existing robotic system new functionalities to deal with robot actions in which there is contact between End Effector and environment and forces are generated due to uncertainty of the environment.

The designated control concept has been the impedance control, aiming to assign a relation between such forces and displacement of End Effector w.r.t. the nominal desired position. During in-contact tasks, the control goal is to make the End Effector behave like a mass-spring-damper system, whose parameters can be arbitrarily specified. This is achieved calculating a correction in the Cartesian space prior to inverse kinematics and through an outer control loop independent from the inner one in charge of position control at joint level.

The correction is calculated by processing Force / Torque sensor readings aquired by the DLR End Effector sensors.
 
 
 
Last update: 27 September 2006