Identifying User Requirements for Digitally Enabling Off-Site Fabrication with Mixed Reality and Gamification

在非现场制造中，民用和建筑基础设施组件设计的定制性质导致了复杂的配置，这些配置很难完全使用自动化设备进行制造。劳动力大量参与大多数任务，使用与传统现场施工非常相似的任务和流程。这导致在现场外生产这些部件的生产率效益最小。混合现实系统的出现，如HoloLens 2及其与安全帽加特林布尔软件的集成，创造了一个机会，通过彻底重新思考工人的任务和流程来解决这个问题，以实现显著的生产力和减少浪费的好处。该项目的目标是利用混合现实系统产生的点云、图像和系统定位数据，自动在构建位置上注册虚拟组件，确定最佳构建顺序，指导工人完成构建，并执行真实的时间尺寸控制，以自我认证构建质量。我们的愿景是建立一个与工人互动的网络物理系统。博士项目将期望学生：a）开发一种自动注册方法，用于将设计模型注册到工作表面，允许手动调整; B）开发近真实的时间分量捕获方法，利用所述设计模型的存在来创建所述对象的初始假设，并且在存在杂乱的情况下将所述真实的世界捕获的几何形状与所述设计模型进行匹配，以及空间误差; c）利用生成对抗网络，仅从设计模型中学习装配过程; d）为装配案例研究开发游戏化+检查原型，以有效指导工人。作为一个起点，MRes组件将涉及广泛的最终用户需求分析，通过与FIBE 2合作伙伴协商，绘制用例的关键活动及其约束条件，沿着优化构建序列的初步尝试。该项目的预期成果是a）装配注册、组件捕获、游戏化和检查的方法; B）概念验证原型; c）出版物和在会议上的介绍; d）开发计划。这项研究将是第一个打破场外制造与基于模型的装配，真实的时间指导和渐进的，自动化检测。它还将产生一个工作原型。

In off-site manufacturing, the bespoke nature of components' design for civil and building infrastructure leads to complex configurations that are difficult to manufacture purely with automation equipment. Labour is heavily involved in most tasks, using tasks and processes very similar to traditional, on site construction. This leads to minimal productivity benefits from producing such components off site. The advent of mixed reality systems, such as HoloLens 2 and its integration with a hardhat plus software by Trimble, creates an opportunity to address this issue off site by completely rethinking the worker tasks and processes to achieve substantial productivity and waste reduction benefits. The project objectives are to leverage point cloud, image and system positioning data produced by mixed reality systems to automatically register virtual components on the build location, determine the optimal build sequence, guide the worker to complete the build, and perform real time dimensional control to self-certify the quality of the build. The vision is to build a cyber-physical system that interacts with the worker.The PhD project will expect from the student to: a) develop an automatic registration method for registering the design model to the work surface, allowing for manual adjustments; b) develop a near real time component capture method, leveraging the presence of the design model to create an initial hypothesis of the objects and matching the real world captured geometry with the design model in the presence of clutter and spatial errors; c) exploit generative adversarial networks to learn the assembly process from the design model alone; and d) develop a gamification + inspection prototype for an assembly case study for effective worker guidance. As a starting point, the MRes component will involve an extensive end-user requirements analysis through consultations with the FIBE2 partners to map the key activities of the use case and their constraints, along with an initial attempt at optimising build sequences.The expected outcomes of the project are a) Methods for assembly registration, component capture, gamification and inspection; b) a proof-of-concept prototype; c) publications and presentations at conferences; and d) an exploitation plan. This research will be the first to disrupt off site manufacturing with model-based assembly, real time guidance and progressive, automated inspection. It will also produce a working prototype.