InVerS Intelligent and Connected Construction Site

In the »InVerS« research project, we are developing AI-based methods to improve safety and resource efficiency on construction sites. At the core of our approach is a central, networked organizational unit that intelligently aggregates sensor data from machinery and construction site infrastructure, identifies safety-critical situations, and uses this information to inform planning decisions for construction processes. By using Digital Twins for building and civil engineering scenarios, we create a safe simulation environment in which we can develop, test, and validate algorithms without having to take risks on real construction sites.

Construction sites are among the most dangerous workplaces. In the EU, approximately 3,000 workplace accidents per 100,000 employees were reported in 2022 (Eurostat). A large proportion of these accidents result from the interaction of heavy machinery with workers who are on foot. In addition to the occupational safety of employees, the protection of uninvolved third parties also plays a central role, for example at inner-city civil engineering sites in close proximity to moving traffic or when interacting with passersby.

At the same time, construction site processes are often inadequately coordinated, which can lead to inefficient workflows, wasted resources, and delays. In this project, we take a comprehensive approach to these challenges by systematically integrating digitalization, connectivity, and AI-based decision support.

The project consists of three phases:

We research the state of the art and design the concept for the central components, sensors, interfaces, and communication channels.
We then develop the methods and test them in a virtual construction site.
Finally, an interactive demonstrator is at the center of our work, where we test and validate the methods.

Communication Makes It Possible: Networking, Interfaces, and the Integration of Real People

To ensure we can seamlessly integrate real people on the construction site later on, we define standardized communication interfaces from the very beginning. This allows us to integrate wearables for construction site personnel as well as retrofit modules (retrofit components) for older machines, which connect the equipment and integrate it directly into the smart construction site management system.

We also include machines whose specific work areas intentionally overlap. At the same time, we detect non-equipped objects or people – such as unauthorized individuals on the site or suppliers without wearables. In such cases, warning alerts are automatically triggered for the affected machines and sent to central security services.

We consolidate all tracked individuals into a central organizational unit and monitor them continuously in real time. Based on motion data, we predict the future movement patterns of machines and people in order to identify potential hazardous situations early on.

To do this, we take into account not only position data but also the physical characteristics of the machines, such as drive systems, steering systems, tire or track models, and performance parameters. This information enables realistic predictions of travel and movement paths.

In this project phase, we are developing two Digital Twins of the construction site – each representing typical scenarios in building construction and civil engineering.

Staying on Top of Things: Boosting Efficiency Through Data-Driven Process Optimization

Our expertise in image processing is also being applied in this project: In the heterogeneous, highly complex construction site environment, AI outperforms traditional methods. Machine learning recognizes objects and complex patterns faster, makes fewer errors, and processes data more robustly.

Comprehensive data collection opens up additional opportunities for us to optimize processes beyond mere safety monitoring. By predicting movement patterns, we can increase energy efficiency, improve material usage, and precisely control the interaction of all components.

One example is asphalt logistics in civil engineering: Pavers, for instance, are supplied by tipper semi-trucks. If the line of trucks is too long, the asphalt cools down and becomes unusable. But if it is too short, downtime occurs. Based on the data collected in the project, our central organizational unit can send specific recommendations for action and control instructions to the relevant personnel to optimize such processes.

Virtual Reality on the Construction Site: Demonstrator and Validation in a Driving Simulator

In the final phase of the project, we are developing a demonstrator. We are using it to make the Digital Twins interactive and integrating them into our robot-based driving simulator, RODOS^®. This step allows us to demonstrate the developed algorithms in a realistic manner and, at the same time, lays the foundation for systematic testing under controlled conditions. Test drivers can interactively immerse themselves in the virtual construction site environment.

Using specially developed scenarios, we comprehensively test the algorithms before deploying them on real construction sites. In addition, we integrate real control units and driver assistance systems into the simulation via hardware-in-the-loop (HiL). We can flexibly adjust environmental conditions such as weather, time of day, or visibility. This allows us to conduct test cycles in a reproducible and efficient manner, for example, to evaluate the performance of different sensor configurations.

Our Expertise at Fraunhofer ITWM

We design and integrate the entire process—from sensor data acquisition to alerting machine operators – into a safe, reproducible simulation environment. In the division »Mathematics for Vehicles, Systems and Production« at Fraunhofer ITWM, we have many years of experience in the key areas of mobility and vehicle development. This includes, among other things:

Operational Stability and Reliability
Energy Efficiency and Operational Optimization
Fleet and Network Management
Driver Assistance Systems and Automated Driving (ADAS/AD)

The methodological focus of our work lies in system simulation, optimization, Artificial Intelligence, and hybrid modeling approaches. This is complemented by extensive expertise in human-machine interaction, sensor fusion, data processing, and AI-based image processing.

We operate our own driving simulator, RODOS^®, and have extensive infrastructure for virtual environments, virtual measurement campaigns (VMC^®), and the institute’s own measurement vehicle, REDAR.

Collaboration and Project Funding

We already have particularly close industrial partnerships in the key areas of commercial vehicles, agricultural machinery, and construction equipment. As a member of the Commercial Vehicle Cluster Southwest (CVC), we are closely networked with relevant research and industry partners. This collaboration forms an important foundation for the successful transfer of project results into practical application.

The project is scheduled to run for two years (October 1, 2025 – September 30, 2027) and is funded under the State of Rhineland-Palatinate’s program for the European Regional Development Fund (ERDF). »InVerS« is part of the »AI Pilot for Mobility« program. It is also embedded in the »Mobility« Transfer Center of the Fraunhofer Center of Excellence for Simulation- and Software-Based Innovation.