Robert Stuart-Smith and Vijay Pawar have an extensive track record of working across construction, manufacturing, and digital sectors in both academia and industry. Their research at University College London’s Autonomous Manufacturing Lab looks at the intersection between autonomous robotic manufacturing and building design; specialising in multi-agent behaviour-based programming that enables humans and machines to collaborate in new ways throughout design, manufacture and construction.  

With the right processes in place, robots can offer improvements in productivity, speed, sustainability and quality by taking on repetitive or dangerous work. While the industry is rapidly adopting an efficient assembly line approach to manufacturing, there are still many environments where heavy machinery and people have to work around each other, such as construction sites. This can result in safety-compromised, overlapping work zones. 

An innovative multi-agent control framework will enable a team of robots to work safely alongside people, by adapting to what is going on around them. This can allow independent and parallel tasks to be carried out in a more productive, fast, sustainable and high-quality manner; whether in offsite manufacturing, onsite construction, or hybrid situations like onsite factories.
Agile mobile ground vehicles (UGVs), aerial vehicles (UAVs), large-scale industrial robot arms, and track and gantry systems will be able to collaborate and collectively undertake tasks beyond the capabilities of each individual machine, such as lifting objects heavier than a single robot's payload capacity.

Social insects such as termites, despite their small size and individual limitations, have the ability to build complex structures through the coordination of many parallel tasks and by working collectively. Inspired by examples in nature, Robert and Vijay wanted to replicate similar flexibility and collaboration in the automation of shared construction environments.

University College London had already demonstrated pilot results in 3D printing with unmanned ground robot vehicles (UGVs) and aerial robots (UAVs) supported by a robot swarm-based control framework.  Building on this, the university's Autonomous Manufacturing Lab sought to develop a collective multi-robot offsite and onsite building manufacturing system.

Construction sites are busy, with lots of tasks happening alongside each other. In order for robots to work safely alongside humans in these environments, they need to be flexible and adaptive to their surroundings and unpredictable events that can occur.

To develop mobile and flexible autonomous manufacturing capabilities, this project explored new methods of robot calibration connected to a digital twin 3D BIM model. This involved looking at robot hardware, control, and metrology-assisted re-positioning to ensure corrective interventions were made when performing manufacturing tasks in large volumes. To explore the adaptive sensing, re-positioning, and path planning of mobile robots carrying out different types of building tasks, a multi-agent sequencing framework was developed to consider new methods using teams of robots working together. This included the development of several agile, high payload, and omnidirectional mobile robot platforms, that can be set up in different configurations to achieve tasks that otherwise pose a risk to humans, such as working at height and manoeuvring heavy or oddly shaped building components into position. This work has been supported by Stuart Robson, Mirko Kovac, Thomas Legleu, Alec Burns, and Durgesh Darekar.

To demonstrate the benefits of autonomous robot data collection on construction sites, a software framework was developed, enabling a mobile ground robot (UGV) to navigate and collect data at regular intervals throughout each day. Working with the AEC Production Control Room a set of time series datasets were produced to compare against and support several parties collaborating throughout the construction process. The feasibility of the robot system operating remotely or unsupervised was also tested. The system has been tested repeatedly on-site in partnership with Mace, providing 360-degree images, video, and high-resolution LiDAR (light detection and ranging) information. AI-based segmentation techniques have also been explored to improve the usability of the data generated using automated methods.  This work was supported by Jacqueline Glass, Grant Mills, Duncan Wilson, Harvey Stedman, and Oliver Dawkins. 

Working with partners is helping to translate this research into demonstrator platforms relevant to the construction industry, by providing the project team with access to specialist equipment, facilities and technical engineering skills:

• Skanska UK and Mace - two of the UK's top construction and development companies
• Atkins and BuroHappold - world-leading engineering practices, experienced with additive manufacturing and innovative methods of construction
• Constructing Excellence - a platform for stimulating, debating, and driving change in the construction sector; engagement with their members shared valuable knowledge

The project team also benefitted from the expertise of the Manufacturing Technology Centre (MTC), part of the Construction Innovation Hub (CIH) funded by the Transforming Construction Challenge. This helped with adoption of the research by industry and access to innovation network partnerships.

Integrating collective autonomous manufacturing with design will open up architectural opportunities. Much like the social insects which inspired them, teams of robots will be able to create complex, geometrical structures without the restraints imposed by scaffolding.

Incorporating collision avoidance capabilities in mobile vehicles, removing the need for human labour in heavy lifting, and reducing the need for people to work at height, addresses three primary causes of fatality and injury.

Delivering the world's first autonomous multi-robot control framework meant integrating multi-robot construction logistics and collective behaviour. In order to provide the robust individual and collective autonomy required to allow a diverse team of robots to operate safely in offsite and onsite environments, the system had to enable different platforms to communicate with each other. The robots in the team shared a global map, objectives, location and state information with high data speeds (low latency) and low error margins. The individual robots also had to have the computational means to make decisions based on what was going on around them.

Increasing use of offsite construction methods and greater integration of robotic autonomy have the potential to transform the construction sector and increase productivity by at least 15%.

Safety
Implementing this research at scale would improve safety in many construction environments. Incorporating collision avoidance capabilities in mobile vehicles, removing the need for human labour in heavy lifting, and reducing the need for people to work at height, addresses three primary causes of fatality and injury.

Uptake
The research team has secured a contract with Homes England to look at using robots to collect data on construction sites.

https://gow.epsrc.ukri.org/NGBOViewGrant.aspx?GrantRef=EP/S031464/1