ACTOR - Automatic coordination of construction actors : Final report

The aim of ACTOR project was to automate the coordination of construction actors with a digital situa-tion picture. In the project, we defined the workflow required for automation. Technical tests were done related to each part of the workflow, including automatic data collection, data storage in a com-mon data lake, data analysis with computer vision, agent-based simulation models and modeling worker decision making, and finally, digital visual management to communicate the results to the work-ers. The developed knowledge can be used by companies to develop commercial solutions. The con-sortium company partners Trimble, Flow Technologies, CarinaFour and Kone have already used some of the research results in their development.
Many solutions developed and tested in the research projects are leading to publications in scientific conferences and journals. The ontology extension work has resulted in one journal paper and two con-ference papers during the project and one additional journal paper has been submitted for review, and one is in preparation. The computer vision tasks will result in at least one journal paper. Automated data collection includes plans for three journal papers. Digital visual management requirements and solutions have been published in two conference papers. The academic output of the project will be 10-15 publications, which is in line with the planned academic deliverables.
However, although all parts of the workflow were tested separately, there were challenges in testing the combination. Developing computer vision is time-consuming, and thus, the project scope was lim-ited to drywall installation. The drywall challenge was solved well, but the other data streams, particu-larly indoor positioning data, were not available for drywall due to a lack of informed consent from drywall workers. Indoor positioning data is critical input information for agent-based simulation, so the simulation model had to be done with another task: furniture installation. This broke the chain from data collection to visual management because all other tasks were based on drywall installation. This limitation should be addressed in future research by ensuring the consent for indoor positioning before selecting task types for computer vision analysis and visual management.
The results presented in this final report and scientific publications offer useful starting points for com-panies developing their systems. Results related to automatic data collection and implementation of the data lake and digital visual management systems are likely to be useful to practitioners. The use of simulation models and digital twins of processes will likely require additional research work before companies are able to use the results. The PI has received research funding from the Research Council of Finland, which allows continuing work on this track.