Built Robotics and Penn xLAB Are Targeting Physical AI for Construction

The University of Pennsylvania’s Safe Autonomous Systems Lab (xLAB) is partnering with Built Robotics to transform construction sites into testing grounds for "physical AI." This collaboration aims to develop a world foundation model designed to manage the coexistence of machines and humans on job sites.

Built Robotics has operated in the field since 2016, focusing on autonomous controls for heavy equipment. The company expanded into the utility-scale solar market in 2023 with its RPD 35, or Robotic Pile Driver. Since its inception, Built has accumulated more than 50,000 hours of operations and deployed at 40+ sites.

The partnership bridges academic research with industrial scale. Rahul Mangharam, principal investigator of xLAB, focuses on safety-critical issues in automating outdoor equipment. By bringing real-world data back to the lab, the collaboration seeks to solve the gap between controlled environment validation and performance under operational conditions.

This is a move toward solving the "edge case" problem in robotics. Built intends to use a new purpose-built, data-collection robot to capture diverse environmental data that standard piling and trenching robots might miss. This includes capturing data on unexpected human behavior, odd body poses, occlusions, and unusual lighting.

For the industry, this represents a shift from simple automation to sophisticated perception. The goal is to build AI models capable of detecting humans by systematically collecting and labeling these difficult edge cases. If successful, the partnership will move physical AI out of the lab and into the high-fidelity mapping environments of active jobsites.

Consider whether the industry can achieve true autonomy without first mastering the unpredictability of human movement in unstructured environments.