Bellman goes Relational (extended abstract)

We introduce REBEL, a relational Bellman update operator that can be used for Markov Decision Processes in - possibly infinite - relational domains. Using REBEL we develop a relational value iteration algorithm.