BIM-Guided Robotic Alignment

We enable construction robotic learning to align components with building information model (BIM) by projecting digital designs directly onto physical workfaces.

Robotic Alignment of Construction Components via a BIM-to-Workface Projection Interface

Mengjun Wang a
Jing Du a
Jiannan Cai b
Shuai Li a
a Department of Civil & Coastal Engineering, University of Florida, Gainesville, FL, United States b School of Civil & Environmental Engineering, and Construction Management, The University of Texas at San Antonio, San Antonio, TX, United States
Paper — Coming soon GitHub — Coming soon

01 / THE PROBLEM & RATIONALE

BIM Designs Are Not Directly Actionable Across Different Robots

OUR RATIONALE

Project BIM-defined targets onto the workface to enable:

  1. Visible Targets

    Robots see where to build.

  2. Learning-Based Alignment

    Robots learn to align using visual feedback.

  3. Shared Reference

    Different robots and workers can use the same targets.

Conceptual robot-specific target delivery: the BIM design target is re-expressed in independent executor coordinate frames.
Conceptual workface spatial interface: the BIM target is materialized at the workface as an observable projected cue; additional executor types illustrate the broader concept.

02 / METHOD

From BIM to Robotic Alignment

We extract construction plans from BIM, project them onto the workface, and use a learned vision-based policy for robots to align construction components.

Architecture from IFC/BIM design target preparation to workface cue placement, local visual observation, corrective motion generation, and final robotic bracket alignment.

03 / PERFORMANCE

Robotic Alignment in Action

The robot aligns brackets with projected BIM targets across three wall appearances using visual feedback.

Ready when visible
Success Rate
80.7%
Mean In-Plane Error
1.65 mm
Trials
135