AI Program: discover LAIghting

Can AI help us automatically generate lighting layout ? Discover VINCI Energies, SDEL, Santerne's solution, one of the twelve projects resulting from Leonard's Artificial Intelligence program (cohort 2021). It's objective? Help VINCI teams become more autonomous by implementing AI use cases, with a high business impact.

Building Information Modeling (BIM) started in the late 1970s with first software tools such as GLIDE, RUCAPS or Reflex. Even more than four decades later, BIM is still pitched as the main disruptor to the construction industry. This term refers to digital representations of a building’s physical and functional properties and includes information about every component that goes into a project.

From BIM to design automation

ArchiCAD, released in 1982, was one of the pioneers in the BIM industry. This software was the first revolution thanks to its ability to provide 2D virtual building representations and to allow collaboration between architects, engineers and other construction professionals. Then, the second revolution was the transition from 2D to 3D modeling and started in 2002 when Autodesk, the company that created AutoCAD, acquired Revit, a BIM software for 3D modeling. It is worth pointing out that for the last decades, BIM profoundly changed the construction industry by focusing on tools that provide richer representations to enable experts from different areas such as ventilation, architecture, etc. to collaborate more efficiently and build more complex building designs. However, the leaders of the CAD (Computer-aided design) industry are convinced that BIM should play a much larger role, and since a couple of years, a new revolution is on its way by shifting the focus of innovations from providing richer representations to automating designs generation.

A new way: generative design

To understand the purpose of this, let’s take a concrete example. Imagine one has to plan the interior lighting layout for an office building, i.e. specify the exact position of each light in a building and its model. By choosing the light model and their positions in the building, performance indicators such as the average amount of light or the uniformity of illumination in the rooms are impacted. For health, security and design purpose, those indicators have to respect certain requirements. In addition to that, the designer must also carefully pick the right light model and reduce its number in the final layout to meet the client’s needs on business criteria such as cost, durability, ecological impact, etc.

To perform this task, a well-trained expert is required. The first challenge is to pick the right light model among the several thousand existing ones. Therefore, the expert needs to have a great knowledge of the lights market and to remain up to date on the latest releases since a lot of innovations occur in this sector. Then, they will rely on his experience to analyze the building and manually propose layouts based on patterns that he encountered during his career. Those designs are then put to the test through a lighting simulation interface to check the quality of various indicators such as the illuminance. When it comes to buildings with dozens of floors and hundreds of squared meters on each of them, the possibilities of layout are almost infinite, thus, it can take some time to get the final output. Regarding the business criteria, the solution is suboptimal and depends a lot on the expert’s skills.

To address such complex tasks in the construction industry and help professionals to build faster and better designs, automatization solutions are emerging and are paving the way for what has come to be called generative design. This approach leverages digitalization of processes and AI-methods to give designers and engineers better insight so they can make faster and more optimized design decisions. Evidence of the interest for this area lies in the $240m acquisition of Spacemaker.io, a startup that has developed AI-supported software for urban development, by Autodesk in 2020, or new initiatives by CAD industry leaders like Autodesk’s Dynamo or McNeel’s Generative design modules.

LAIghting, a solution to automate the lighting layout generation

The CAD industry being strategic for VINCI Energies, the entity decided to step into that breach. With the technical skills of VINCI Energies France IDF, the expertise of SDEL and Santerne in the lighting industry and the support of the Leonard AI program, a collaboration emerged to launch LAIghting, a solution to automate the lighting layout generation.

LAIghting is a generative design-driven solution that leverages data science methods to automate and optimize lighting layout design generation. Inspired by the dynamics of swarm movements, the solution relies on what is called the PSO (Particle Swarm Optimization) model to find an optimal lighting layout. The algorithm takes as input the building’s blueprints, then the expert indicates to the model the client’s requirements and some indicators of interest such as the cost or the ecological impact. The solution automatically proposes an optimized layout according to the constraints and the selected criteria, as well as the best light model among a pre-selected sample.

To improve the solution even further, an on-going project has been launched to build a centralized database with all the technical and commercial light specifications. The database enables LAIghting to enlarge its light selection catalog to obtain better layouts, but it also can be used by other generative design initiatives and therefore push even further innovation capabilities of VINCI Energies France in the sector. For the next period, the main goal is to industrialize a first version of the solution through some pilot clients and this achievement constitutes a step forward in designing greater interior offices while respecting what one has envisioned for its building in term of costs, ecological impact or other criteria of interest.

This article is part of a series involving the participants of the AI program by Leonard. The program has been specifically designed to accelerate the adoption of AI technologies within the VINCI group. It consists in a five-month incubation period where selected VINCI collaborators follow a learning by doing process where they codevelop an AI-based use case under the coaching and mentoring of the Leonard Team and Eleven consultants.

> More about Leonard’s AI program 

What do we do?

The VINCI Group created Leonard to tackle the challenges posed by the transformation of regions and lifestyles. Our goal is to unite a community of key stakeholders in order to build the city of the future together.