Modelling of Learning and Perception

Centre of Excellence in Computational Complex Systems Research


LCEBOT: Cognitive and Interactive Robot

The goal of the project is to develop cognitive skills for user interfaces that are unobtrusive, acceptable and easy to use. The development platform is an autonomous robot with speech synthesis, vision and stereo hearing, and an attached animated talking head. The system will be able to learn the environment without being provided with exact CAD model of the space, by active (curiosity driven) exploration and learning from incomplete and sparsely collected data. Similarly, the system will be able to learn the typical behaviour of people in the environment, for example to use similar gestures as humans use for understanding and indicating the direction of motion. This ability is necessary for autonomous agents moving among people.

Learning and inference will be based on Bayesian approach, by representing uncertainty in observed data and learned models by probabilities, and by using numerical Monte Carlo techniques to compute the probabilities of the goal attributes given the observations. Both learning, inference and prediction can be carried out by the same principle.

Although the developed robotic system is directly targeted for studying human-computer interaction with autonomous agents, the same methods can be used for modelling the environment and learning and predicting the user requirements in many other modes of human-computer interaction (e.g., in pervasive computing).



The hardware and basic control software have been contracted from the Intelligent Machines and Special Robotics Institute, IMSRI, HUT.


The research team consists of Prof. Jouko Lampinen, Acad.Prof. Mikko Sams, Dr.Tech Harri Valpola, Dr.Tech Aki Vehtari, Dr.Tech. Michael Frydrych, M.Sc Ilkka Kalliomäki, M.Sc Timo Kostiainen, M.Sc Simo Särkkä, and M.Sc Toni Tamminen.

Summer workers: Juho Kettunen, Jaakko Riihimäki, Juho Sundquist, and Simon Finnegan.

Contact Person

Jouko Lampinen, Responsible Manager
Tel. (09) 451 4827, gsm: 050 560 4827