Reasoning With Scene Graphs for Robot Planning Under Partial Observability

Robot planning in partially observable domains is difficult, because a robot needs to estimate the current state and plan actions at same time. When domain includes many objects, reasoning about objects their relationships makes even more difficult. In this letter, we develop an algorithm called scene analysis for planning (SARP) that enables robots reason with visual contextual information toward achieving long-term goals under uncertainty. SARP constructs scene graphs, factored representation of relations, using images captured from different positions, reasons them enable context-aware partial observability. Experiments have been conducted multiple 3D environments simulation, dataset collected by real robot. comparison standard methods, target search domain, improves both efficiency accuracy task completion.