Collaborative Research: FW-HTF: Integrating Cognitive Science and Intelligent Systems to Enhance Geoscience Practice

人类技术前沿的未来工作（FW-HTF）是NSF宣布的10个未来投资新想法之一。FW-HTF跨董事会计划旨在通过支持融合研究来应对不断变化的就业和工作环境的挑战和机遇。这个奖项实现了这个目标的一部分。该项目将为支持未来的地质工作者做出重大贡献。了解地质学家如何推理，计划收集新数据，考虑三维空间关系和评估不确定性对于支持科学家解决应用问题（如自然资源勘探）至关重要。该项目将加强现有的地质工作，使用无人驾驶机器人收集数据。 无人机允许进入世界上的重要地区，这些地区太危险，无法亲自进入，也无法从卫星或飞机上看到。 该项目将使用机器学习将专家知识融入无人机飞行中，以支持有效的自主数据收集。 这些数据将使人们对一个重要的断层系统有更好的地质认识。该项目的发现将提高对数量不确定性的理解，从而提高我们对地震的理解和石油工人的分析。 了解专家地质学家如何推理将为在自然环境中工作的人类机器人团队提供新的勘探和绘图策略。跨学科团队的合作努力将推动认知科学，地质学和机器学习领域的发展。认知科学、机器人技术和地质学的整合将开发出与人类自主系统团队一起进行实地工作的新方法，这些方法比任何人类或自主系统单独行动都更快、更有效。该项目将描述专家对3D关系和不确定性的空间推理，因为地质学家收集数据以开发对新领域的3D理解，对未来的观测进行预测，并构建地质模型。 关于3D结构的推理错误将用于开发专家不确定性的定量模型。这些模型将被用来帮助明确可视化专家的不确定性，并构建机器人导航的成本函数。成本函数将包括获取科学价值的指标。该项目将开发无人机勘探和测绘的新方法，包括对地质学家感兴趣的特征进行机器学习。无人机将自主探索和绘制峡谷环境中的天然岩层，以支持和加快野外地质学家的数据收集和解释工作。 该项目将研究加州麦加山地区的结构地质，该地区是圣安德烈亚斯断层系统的一个暴露部分。 无人机将收集有关表面特征的数据，以绘制地下结构的地图。融入认知科学的机器人设计将采用成功的专家策略，并专注于专家可能出错的领域，以在导航计划中优先探索这些领域。所提出的策略将使三维表面重建的峡谷表面。它们还将使人们更好地了解如何加强规划和专家在收集科学上重要的数据方面的即时决策。 该项目的基础工作旨在发展一种跨学科的理解，即地质学家如何随着时间的推移对一个地区建立科学的理解。该奖项还旨在为人类-机器人团队设计自主探索策略，并测试新方法，以支持在何处收集数据以最大限度地发挥科学影响的顺序决策。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

The Future of Work at the Human-Technology Frontier (FW-HTF) is one of 10 new Big Ideas for Future Investment announced by NSF. The FW-HTF cross-directorate program aims to respond to the challenges and opportunities of the changing landscape of jobs and work by supporting convergent research. This award fulfills part of that aim. This project will make a significant contribution toward the support of future workers in geology. Understanding how geologists reason, plan to collect new data, consider three-dimensional spatial relations, and evaluate uncertainty are critically important for supporting scientists working on applied problems, such as natural resource exploration. This project will enhance existing efforts in geology to collect data using robot drones. Drones allow access to important areas of the world too dangerous to access in person and not visible from satellite or plane. The project will use machine learning to incorporate expert knowledge into drone flights to support effective autonomous data collection. The data will yield improved geological understanding of an important fault system. Findings from the project will improve understanding of uncertainty in volumes and thus improve our understanding of earthquakes and the analyses of petroleum workers. Understanding how expert geologists reason will support new exploration and mapping strategies for human-robot teams working in natural environments. The collaborative efforts of the interdisciplinary team will advance the fields of cognitive science, geology, and machine learning. The integration of cognitive science, robotics, and geology will develop new approaches to field work with human-autonomous systems teams that are faster and more effective than any either human or autonomous system would be acting alone. The project will characterize expert spatial reasoning about 3D relations and uncertainty as geologists collect data to develop a 3D understanding of a new field area, make predictions about future observations, and construct geological models. Errors in reasoning about 3D structures will be used to develop quantitative models of expert uncertainty. These models will be used to help explicitly visualize uncertainty for the experts and to construct cost functions for the robot navigation. The cost functions will include metrics that capture scientific value. The project will develop new approaches to drone exploration and mapping, including machine learning of features of interest to geologists. Drones will autonomously explore and map natural rock formations in canyon environments to support and speed up the data collection and interpretation efforts of field geologists. The project will study the structural geology of the Mecca Hills area of California, a well exposed portion of the San Andreas fault system. Robot drones will collect data about surface features to develop maps of subsurface structures. The cognitive science-infused robot design will employ successful expert strategies and focus on areas where experts are likely to make errors to prioritize exploration of those areas in navigation plans. The proposed strategies will enable 3D surface reconstruction of canyon surfaces. They will also enable better understanding of how to enhance planning and on-the-fly decision making of experts for collecting scientifically important data. The project's foundational work aims to develop an interdisciplinary understanding of how geologists build a scientific understanding of a region over time. It also aims to design autonomous exploration strategies for human-robot teams, and test new ways to support the sequential decisions about where to collect data to maximize scientific impact.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.