This paper describes an architecture that combines the complementary strengths of probabilistic graphical models and declarative programming to enable robots to represent and reason with qualitative and quantitative descriptions of uncertainty and domain knowledge. The architecture comprises three components: a controller, a logician and an executor. The logician uses a coarse-resolution, qualitative model of the world and logical reasoning to plan a sequence of abstract actions for an assigned goal. For each abstract action to be executed, the executor uses probabilistic information at finer granularity and probabilistic algorithms to execute a sequence of concrete actions, and to report observations of the environment. The controller dispatches goals and relevant information to the logician and executor, collects and processes information from these components, refines the coarse-resolution description to obtain the probabilistic information at finer granularity for each abstract action, and computes the policy for executing the abstract action using probabilistic algorithms. The architecture is evaluated in simulation and on a mobile robot moving objects in an indoor domain, to show that it supports reasoning with violation of defaults, noisy observations and unreliable actions, in complex domains.

本文介绍了一种机器人体系结构，它结合了概率图形模型和声明性编程的互补优势，用于表示和处理基于逻辑和概率的不确定性和领域知识。它采用行动语言描述了两个粒度水平上的紧密耦合的过渡图，通过求解部分可观察的马尔可夫决策过程，支持在复杂领域中处理默认违规、嘈杂观察和不可靠行为的推理和规划。

REBA: 机器人知识表示和推理的基于细化的体系结构