.gif "TU/e"){kind=small}
Short biography
I started my PhD in robotics in 1989, after obtaining master degrees in mathematical physics, computer science, and mechatronics. The context of my research was provided by the realtime sensor-based robot control research activities of Hendrik “Rik” van Brussel (who pioneered robotics in Leuven in 1975, and holonomic manufacturing systems from the 1990s) and Joris De Schutter (who identified the essential relations between “compliant motion” task specification, hybrid force/velocity control and the mechanical compliance properties of robot, tool and environment). Especially Joris was instrumental in my research towards introducing fundamental structures into sensor-based robot control: invariance of solutions with respect to changes in reference frames and physical units; redundancy and constraint resolution; task specifications distributed over several locations on robot, tools and environment, formally represented as constrained optimization problems; integration of uncertainty in task specification and control, in the form of, both, Bayesian probability theory and multi-hypothesis world models.
As an early fulltime adopter of Linux from 1994 on, and after postdoc stays at the University of Pennsylvania (1996) and Stanford University (1999), I decided to make software engineering for robotics one of my research focuses, and vendor-neutral open standards as one of my technology advocacy focuses. The natural outcome of both is quite some seminal open source software for realtime, distributed and heterogeneous system-of-system robotics applications.
Since 2010, I started research on formal knowledge representation in robotics, in the form of hierarchically and heterarchically interdependent property graph models. The aim of this effort is to design robots that can exploit these knowledge graphs themselves (and not only their human software developers) to discover each other, to find out with whom to cooperate on tasks that can (or have to) be shared, and to explain each of their decisions, to themselves, to their peer robots, and to human stakeholders (developers, users, certifiers and regulators).
The major lesson I learned after 30+ years of making robotics applications is that overall progress in robotics has slowed down tremendously, with even, in times of the “deep learning” hype, an amazing amount of “forgotten insights”! in the community mainstream. The major cause of this evolution is that we fail to give our robots enough high-level context, through which to integrate the essential five aspects of any robotic application: world models, formal task specifications, perception, control and situational aware skills. These aspects must be tackled, at the same time and in an integrated way, at a dozen or so “levels of abstraction”, coupled by “5Cs” software architectures. (That is, the Composition of Computation, Communication, Coordination and Configuration).
Explaining that integration is something that goes beyond the limits of conference or journal publications, which is why I started a longterm dissemination project, in the form of an online “work in progress” book about the above-mentioned “5Cs”: Building blocks for complicated and situational aware robotic and cyber-physical systems.
