In post-earthquake surveys, it is difficult (and often infeasible) to observe and quantify displacements beyond line-of-sight (LOS), given seismic force-resisting and gravity systems exist completely or partially within a building’s enclosure. To overcome this limitation, we develop a novel framework that generalizes graph-based state estimation towards structural joint localization via engineered landmarks. These landmarks provide an indirect means to estimate residual displacements where direct LOS is unavailable. Within our framework, engineered landmarks define topologies of uniquely identifiable landmarks that are either visible or non-visible to a robot performing simultaneous localization and mapping (SLAM). Within the SLAM approach, factors encoding robot odometry and robot-to-visible landmark measurements are formulated for the cases of wireless sensing and fiducial object detection and tracking. Visible landmarks are rigidly attached to non-visible landmark subsets for each engineered landmark, where the complete set of non-visible landmarks form globally rigid and localizable connectivity graphs via range-based factors. Complimentary subsets of non-visible landmarks are embedded within the base structure and uniquely define joint pose via geometric factors. All factors are unified within a common graph to solve for the maximum a posteriori estimate of robot, landmark, and joint states via nonlinear least squares optimization. To demonstrate the applicability of our approach, we apply the Monte Carlo method over a parameterization of system noise to calculate residual joint pose error distributions, maximum average inter-story drift ratios, and related summary statistics for a 19-story nonlinear structural model. By performing nonlinear time history analyses over sets of service-level and maximum considered earthquakes, our parametric study gives insight into our method’s application towards post-earthquake building evaluation in non-LOS conditions.

在震后调查中，鉴于抗震力和重力系统完全或部分存在于建筑物的围护结构内，很难（而且通常是不可行的）观察和量化超出视线 (LOS) 的位移。为了克服这个限制，我们开发了一个新的框架，通过工程地标将基于图形的状态估计推广到结构联合定位。这些地标提供了一种间接方法来估计直接 LOS 不可用时的残余位移。在我们的框架内，工程地标定义了独特可识别地标的拓扑结构，这些地标对于执行同步定位和映射 (SLAM) 的机器人来说是可见的还是不可见的。在 SLAM 方法中，针对无线传感和基准物体检测和跟踪的情况，制定了编码机器人里程计和机器人到可见地标测量的因素。可见地标严格依附于每个工程地标的不可见地标子集，其中整套不可见地标通过基于范围的因素形成全局刚性和可本地化的连接图。不可见地标的互补子集嵌入基础结构中，并通过几何因素唯一定义关节姿势。所有因素都统一在一个公共图形中以求解最大值 其中整套不可见地标通过基于范围的因素形成全局刚性和可本地化的连接图。不可见地标的互补子集嵌入基础结构中，并通过几何因素唯一定义关节姿势。所有因素都统一在一个公共图形中以求解最大值 其中整套不可见地标通过基于范围的因素形成全局刚性和可本地化的连接图。不可见地标的互补子集嵌入基础结构中，并通过几何因素唯一定义关节姿势。所有因素都统一在一个公共图形中以求解最大值通过非线性最小二乘优化对机器人、地标和关节状态进行后验估计。为了证明我们方法的适用性，我们将蒙特卡罗方法应用于系统噪声的参数化，以计算 19 层非线性结构模型的残余关节位姿误差分布、最大平均层间漂移比和相关汇总统计数据。通过对一组服务水平和最大考虑地震进行非线性时间历史分析，我们的参数研究深入了解了我们的方法在非 LOS 条件下对震后建筑评估的应用。