Since 2007 we have implemented different processes for inferring knowledge in real robots. The two main lines of research in this topic are semantic maps and sensor recovery. In the former, we have included semantic information into robotic operations such as task planning and navigation in order to perform more efficiently (by reducing computational cost…

In 2005 we started to work on the application of Reinforcement Learning to real robots. About a decade later we extended that research to the decision making process in general (not necessarily involving learning). All in all, we have studied RL in diverse contexts within robotics: robot task planning, onboard computer efficiency, robot reflex learning,…

Since the early 2000s we have addressed task planning for mobile robots using different cognitive processes, such as abstraction or semantics. The results have demonstrated improved computational cost and robustness.

We are working since 1997 on implementing abstraction into a number of diverse robotics tasks, such as world modelling, task planning, SLAM or sensor inference. We have proven in all these contexts how this cognitive process turns exponential complexity into polinomial one in suitable situations. We have also developed the first theoretical model of multiple…