Building a robust and trustworthy (semi-)autonomous agent requires us to build a consistent picture of the state of the world based on the data received from some perception module.

In this project we explore the use of modelling formalisms like Answer Set Programming (ASP) and Constraint ASP (such as the one implemented in the s(CASP) system)  to encode the background knowledge and commonsense reasoning required for our agent to succeed in tasks such as those posed by the Animal AI-Testbed. This will require us to  explore the use of probabilistic-aware, non-monotonic rules, facts and constraints that can be used by the agent to build a world state (i.e., a static snapshot of the environment at some time T), while also maintaining a historical view formed by previous snapshots. Each world state must contain a consistent set of objects, their information and the associated relations among these objects.

Students in this project will collaborate closely with those in projects “Efficient exploration of consistent worlds” and “Solving Automata using ASP/Minizinc?”