Monday December 16, 2013 at 16h30 in Celestijnenlaan 200A (auditorium 00.225)
DCPF: A Relational Framework for State Estimation in Hybrid Temporal Models
by Davide Nitti (PhD student DTAI)
Most of the literature in declarative artificial intelligence regarding autonomous agents assumes a discrete world with relations between objects and discrete states without caring about low-level noisy information. On the other hand, most of the literature in robotics and vision handles low-level information (sensors and actuators) without really taking into account high-level relational information (e.g. background knowledge). The gap between this high and low level of reasoning is currently an unsolved problem which receives insufficient attention from both scientific communities involved. We try to bridge this gap with a relational framework for hybrid temporal models called Distributional Clauses Particle Filter (DCPF). The framework is based on a dynamic version of Distributional Clauses, an extension of Sato's distribution semantics. Several probabilistic logic languages are based on Sato's distribution semantics, including Problog, PRISM and CPL. Nonetheless, Distributional Clauses is one of the few probabilistic logic languages that supports continuous distributions. We describe representation, inference, online parameter learning and several object-tracking applications.