An Experience-Based Interface for Abstracting the Motion Control of Kinematically Complex Robots
In Proceedings of ASTRA 2015, (ASTRA-2015), 11.5.-13.5.2015, Noordwijk, ASTRA, 2015.
In order to provide higher mobility and to assist humans
in building up infrastructure in future extraterrestrial
space missions, kinematically complex robots are
needed. One key challenge which needs to be addressed
is to handle their complex motion control and to make
use of their high potential. Utilizing the possibility to
achieve various actions even in different ways by tuning
manually numerous parameters of the motion control can
be very demanding and even unmanageable when also
taking communication delay into account.
Thus, the proposed experience-based interface is encapsulating
the motion control of complex robots by autonomously
mapping application-specific action parameters
to robot-specific motion control parameters depending
on the current context. Therefore, the robot is using
experiences collected from previously executed behaviors.
Apart from acquiring experiences during operation
of the real robot, they can also be collected in simulation.
The possibility to test in low gravity environments
makes the latter a valuable tool for increasing the robot’s
knowledge base for space missions.
The experiments in this paper show that reconfiguring
the motion control can be beneficial and that in simulation
optimized behaviors can easily be integrated in the
experience-based control interface to improve the performance
of a robot. In addition, the transferability from
simulation to the real system is shown.
Key words: kinematically complex robots, experiencedbased
control, behavior adaptation.