Learning Multi-Agent Control with OROCOS

In previous research projects, we have used a number of different techniques / methods that are applicable in advanced control of mechatronic systems, i.e.: • Learning Feed-Forward Control; see e.g. http://www.ce.utwente.nl/rtweb/publications/2008/ pdf-files/Cuong_thesis.pdf, http://www.ce.utwente.nl/rtweb/publications/2004/pdf-files/krf/ and http://www.ce.utwente.nl/rtweb/publications/2000/pdf-files/vts/ • Multi-Agent Control Systems; see e.g. http://www.ce.utwente.nl/rtweb/publications/Msc2003/pdf-files/010CE2003_Eglence.pdf and http://www.ce.utwente.nl/rtweb/publications/2001/pdf-files/AJN van Breemen.pdf • OROCOS (Open Realtime Control Services and Open Robot Control Software) running on top of Xenomai/RTAI; see www.orocos.org , www.xenomai.org and www.rtai.org Currently, efforts are made to integrate these methods so as to obtain a suitable framework for advanced control of mechatronic systems. Specifically, we envision Multi-Agent Control Systems to become Tasks in OROCOS, and LFFC as a pattern for incorporation in a Multi-Agent Control System, with the specific property that learning is done asynchronously, in a separate non-realtime Task. Goal of the project is to evaluate the feasibility of the proposed integration. This is to be done by designing and implementing a simple Multi-Agent Control System with path generation and PID control in OROCOS and subsequently adding a Learning Feed-Forward Component that can learn on-line