Graphical Models Actions Bell p(xlap Observations Observable 2 hidden state Want to estimate p(Xi x, ax), p(zixi)instead of 区X1x2,…,Xx Topological Maps Idea: build map as graph of sensor signatures and control laws connecting Extracted Topology nodes Actions are control laws Observations are .? p()={0,1} Metric map layered on top Layer geometric information on topologyGraphical Models States x1 x2 p(xj |ai , xi ) Z2 b Beliefs 1 Observations Z1 a Actions 1 p(zj |xi ) b2 Z2 Hidden Observable ● Want to estimate p(xi |xj ,ak), p(zi |xj ) instead of {x1,x2,...,xT} Topological Maps ● Idea: build map as graph of sensor signatures and control laws connecting nodes ● Actions are control laws ● Observations are...? ● p(•|•) = {0, 1} ● Layer geometric information on topology Extracted Topology Metric map layered on top