Correct High-level Robot Behavior in Environments with Unexpected Events
Correct High-level Robot Behavior in Environments with Unexpected Events
复制标题
在发生意外事件的环境中纠正高级机器人行为
DOI：
10.15607/rss.2014.x.012
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
H. Kress
中科院分区：
文献类型：
--
作者：
K. W. Wong;Rüdiger Ehlers;H. Kress
从高级别规范合成正确的结构机器人控制器具有在不同环境下提供有保证的机器人行为的优点。通常，当合成这样的控制器时，用户对环境的行为所做的假设（如果有的话）被合并到所得到的控制器中。然而，在实践中，环境假设可能是未知的用户，从而防止合成的应用。即使环境假设是可用的，它们也可能由于建模错误或不可预见的异常操作条件而在机器人执行期间不成立。在本文中，我们解决这两个问题。我们提出了一种方法，用于合成控制器，包括内置的恢复过渡，使机器人在环境假设违反的情况下，尽可能地朝着目标前进。此外，我们提出了一个自动增强规范与环境的假设，从机器人的观察在运行时计算的过程。我们从一组候选假设开始，只要在运行时被违反，就会更新这些假设。
Synthesis of correct-by-construction robot controllers from high-level specifications has the advantage of providing guaranteed robot behavior under different environments. Typically, when such controllers are synthesized, assumptions that the user makes about the behavior of the environment, if any, are incorporated into the resulting controller. In practice, however, the environment assumptions may be unknown to the user, thus preventing the application of synthesis. Even if environment assumptions are available, they may not hold during the robot’s execution due to modeling errors or unforeseen anomalous operating conditions. In this paper, we address both of these problems. We present an approach for synthesizing controllers that include built-in recovery transitions, enabling the robot to make progress towards its goals in the event of environment assumption violation, whenever possible. Furthermore, we present a process for automatically augmenting a specification with environment assumptions that are computed from the robot’s observations at runtime. We start with a set of candidate assumptions that is updated whenever violated at runtime.