This paper presents Workspace-based Connectivity Oracle (WCO), a dynamic sampling strategy for probabilistic roadmap planning. WCO uses both domain knowledge—specifically, workspace geometry—and sampling history to construct dynamic sampling distributions. It is composed of many component samplers, each based on a geometric feature of a robot. A component sampler updates its distribution, using information from the workspace geometry and the current state of the roadmap being constructed. These component samplers are combined through the adaptive hybrid sampling approach, based on their sampling histories. In the tests on rigid and articulated robots in 2-D and 3-D workspaces, WCO showed strong performance, compared with sampling strategies that use dynamic sampling or workspace information alone.