Research in Artificial Intelligence and applications
Research at CRIL deals with the design of intelligent autonomous systems. Depending on the available information, those systems should be able to take reasonable decisions to reach given goals. As such, some form of reasoning is needed. The main difficulties to build such systems are diverse. First, the available information is usually heterogeneous and imperfect. That information contains knowledge and beliefs about the state of the world in which the agents evolve (for example, physics law, but also data gathered from sensors more or less reliable). It contains the information about the other agents found in that world, the description of available actions and their effects, the preferences of the agents on the state of the world or the actions to perform. The imperfection of the available information has several aspects (all correlated): incompleteness, incoherence, contextualization, etc.
The kind of inference necessary to simulate an “intelligent” behavior are multiple. The succinctness of the representation languages often make inference and decision making computationally untractable in the worst case. It is thus important to identify the sources of the underlying complexity to tackle them when possible, by developing specific algorithms efficient in practice, or using approximating or compilation methods. CRIL develops its research activities following two main axes: on the one hand, processing imperfect, dynamic contextual and multi-source information; on the other hand, algorithms for inference and decision making.