The Construction 5.0 Revolution: Cyber-Physical Digital Twins and Intelligent Robotics in Industrial Facade Execution

The Architecture, Engineering, and Construction (AEC) industry faces long-standing challenges regarding stagnant productivity and the urgent need to enhance workplace safety. In response, Construction 5.0 has emerged as a design paradigm focusing on the harmonious, intelligent, and safe collaboration between human workers and robotic systems. Under the framework of the 6th Portuguese Conference on Building Information Modelling (ptBIM 2026) held at the Faculty of Engineering of the University of Porto, a pioneering research paper was presented under the title "Digital Platform to Support the Execution of Facades in Industrial Buildings with the Support of Robotics and Artificial Intelligence".

Developed within the scope of the CRIARTE project (Construção com Robótica Inteligente e Arquitetura Revolucionária de Tecnologias Emergentes), this innovative approach bridges the semantic and geometric discontinuity between the static planned digital model (As-Design in BIM) and the dynamic, unpredictable reality of an active construction site (As-Is). The proposed architecture establishes a unified cyber-physical system orchestrating three distinct Digital Twins (DTs): the Asset, the Human, and the Robot, seamlessly integrated through the ROS2 (Robot Operating System 2) middleware and the Unity engine.

Keywords: Digital Twins, Construction 5.0, ROS2, Sensor Fusion, Collaborative Safety.

Digital Platform to Support the Execution of Facades in Industrial Buildings with the Support of Robotics and Artificial Intelligence - PTBIM 26

1. The Asset Digital Twin: Closing the Geometric and Environmental Loop

The Asset Digital Twin serves as the virtual replication of the evolving physical structure. Its architecture operates across three main monitoring pillars driven by advanced Sensor Fusion:

Geometric Verification

Using a heavy manipulator robot (mounted on a Manitou MRT 2260 mobile platform) as the primary data acquisition vehicle, the system collects environmental data through high-density Leishen CH64W LiDAR sensors and StereoLabs ZED 2i stereoscopic cameras. To process the immense volume of raw spatial data, the captured point clouds undergo voxel downsampling and noise filtering. Registration with the original BIM model is executed via the classic Iterative Closest Point (ICP) algorithm combined with Deep Learning models. This enables the calculation of metric deviations, panel coverage rates, and point-to-surface distances in real time, immediately flagging structural misalignments.

Anomaly Mapping

To identify structural degradation during the construction process, Unmanned Aerial Vehicles (UAVs) capture high-resolution, georeferenced images of the building's envelope. This imagery feeds a real-time YOLO (You Only Look Once) machine learning pipeline tasked with detecting and segmenting surface damages such as corrosion or mechanical impact deformities. Once a 2D defect mask is generated, a Ray Casting algorithm projects virtual rays from the camera’s 3D optical center until they intersect with the asset’s virtual mesh, stamping the pathologies directly onto the Digital Twin database for metric tracking.

Environmental Assessment

Operational safety on a real-world construction site depends heavily on atmospheric stability. An Ecowitt WN90LP weather station continuously monitors wind velocity, ambient temperature, and precipitation. Governed by a Go/Reduce/Stop decision matrix structured in strict compliance with ISO 10218 and ISO/TS 15066 collaborative robotics standards, the system dynamically restricts high-risk maneuvers (such as lifting heavy sandwich panels). In the event of communication failures or critical weather violations, the robot independently triggers an automated, fail-safe emergency stop.

Asset Digital Twin

2. The Human Digital Twin: Proactive Safety and Ergonomic Analytics

Diverging from traditional safety methods that reactively log accidents, the Human Digital Twin prioritizes proactive prevention by tracking the physiological and contextual state of site workers.

Operating in strict compliance with the General Data Protection Regulation (GDPR) via data anonymization and informed consent protocols, workers are equipped with biometric sensor packages (such as Polar H10 chest straps and Empatica EmbracePlus wristbands). These wearables monitor vital metrics including heart rate, skin temperature, electrodermal activity (EDA), and heart rate variability (HRV)—specifically extracting the time-domain parameters SDNN and RMSSD.

These biometric streams are structured into JSON data packets and cross-referenced with spatial tracking data from site-bound LiDAR sensors that capture the real-time position and joint pose of the personnel. Recurrent neural networks using LSTM (Long Short-Term Memory) architectures analyze these concurrent inputs to differentiate between standard physical fatigue and a sharp stress response caused by immediate hazard exposure—such as an inadvertent intrusion into the dynamic workspace of the robotic arm. When high cognitive load or imminent collision risks are inferred, the system broadcasts immediate visual alert layers directly to the site manager's Microsoft HoloLens 2 Augmented Reality (AR) headset, enabling instant human intervention.

3. The Robot Digital Twin: Predictive Motion Planning and Hydraulic Controls

The Robot Digital Twin simulates, monitors, and optimizes the kinematic and dynamic behaviors of the machinery. Its backend architecture is segmented into seven core logical layers ranging from low-level data acquisition (LiDAR, GNSS, IMU) to high-level spatial mapping using SLAM (Simultaneous Localization and Mapping) algorithms.

The Goal Identification (Goal ID) layer digests static assembly parameters from the BIM database—such as panel sequencing, tolerances, and optimal suction grip points—and translates them into active robotic targets. Subsequently, the MoveIt2 motion planning framework generates 3D collision-free trajectories within a localized planning scene that is continuously updated by real-time field perception.

To achieve millimeter-level positioning precision while manipulating heavy panels using non-linear hydraulic actuators, the control layer employs highly adaptive control strategies:

1. Baseline Control: Standard kinematic loop for position, velocity, and acceleration profiling.

2. Fuzzy-PID Controllers: Combine traditional PID robustness with fuzzy logic to absorb physical site uncertainties, structural elasticity, and joint backlash.

3. Fractional-Order Controllers: Offer supplementary mathematical parameters to manage complex fluid dynamics and pressure saturation curves within the hydraulic lines.

4. Deep Learning Model Predictive Control (MPC): Operates in parallel to anticipate mechanical latencies and actuator lag, applying mathematical corrections to the trajectory before any deviation physically materializes.

5. 
   

4. The Integrating Digital Platform: Interoperability and Unified 3D Dashboard

The primary structural achievement detailed in the research is the deployment of the Unified Digital Platform, which acts as the nervous system of the cyber-physical architecture. Data fragmentation—traditionally the largest barrier to digitalization in civil engineering—is resolved by employing ROS2 as a standardized middleware core. High-throughput data streams (Asset point clouds, Human biometrics, and Robot joint telemetries) are transmitted via standardized ROS2 topics managed under strict Quality of Service (QoS) and data compression profiles to maintain low network latency.

The human-machine interface (HMI) is developed inside the Unity engine, functioning as a completely decoupled node through the ROS-TCP Connector package. Unity renders a fully immersive 3D dashboard that overlays critical operational dimensions into a singular view: the geometric assembly progress of the facade, calculated robotic paths, ray-casted anomaly maps, and dynamic human safety zones. Furthermore, it supports full playback capabilities using rosbag archives, facilitating post-operational audits and data-driven optimization of future assembly workflows.

Conclusion and Future Horizons

The framework validated under the title "Digital Platform to Support the Execution of Facades..." represents a major technological leap beyond fragmented field tools. By treating the physical structure, the human workforce, and heavy robotics as a unified, deeply intertwined cyber-physical ecosystem, Construction 5.0 transitions from an abstract theory into an operational reality.

The consortium's next development phases center on transitioning this conceptual architecture into continuous field validation on active industrial sites. By monitoring key performance indicators (KPIs)—including spatial registration accuracy and end-to-end telemetry latency—this digital platform establishes a clear engineering pathway toward a safe, automated, and industrialized construction sector.