Logic-Based Subsumption Architecture: Empirical Evaluation (Status Report)

We describe an implementation of the Logicbased Subsumption Architecture (LSA) [Amir and Maynard-Reid, 1999], an AI architecture inspired by Brooks’ subsumption architecture but based on logic. In this architecture, layers of control are axiomatized in First-Order Logic (FOL) and independent theorem provers derive each layer’s goals given its inputs. Default assumptions in the lower layers implement subsumption by higher layers. These assumptions are nonmonotonically retracted when higher layers draw new conclusions. We describe four layers designed for the task of robot control and experiments that empirically show the feasibility of using fully expressive FOL theorem provers for robot control with our architecture.