Descripció del projecte
Approximately 20% of construction costs are wrapped up in fixing errors. This project is aimed at providing robotic solutions to minimize the rework by automating progress monitoring. There is a need for high-resolution, perfectly aligned comparisons of the digital models made by architects to the on-the-ground build site to help build managers keep close track of progress and spot problems before they can scale into costly expenses. Leveraging LiDAR and autonomous vehicle technology, Scaled Robotics is interested in building mobile robots that navigate and build maps of construction sites by fusing images, video and data captured by its robots. This task requires a perfectly calibrated robot that is robust for use in harsh environments such as a construction setting.
This project is aimed at intrinsic and extrinsic multi-sensor calibration and probabilistic state estimation on the basis of sensor fusion theory. The fusion of data and the calibration of proprioreceptive sensors, such as inertial measurement units (IMUs), and exteroceptive sensors, such as LiDARs, cameras or other range sensing devices will be considered in this work. By intrinsic calibration we refer to the estimation of all intrinsic sensor parameter such as acceleration biases on IMUs or focal length on a camera; or to the kinematic parameters of the robot. By extrinsics we refer to the relative positioning of the various sensor reference frames with respect to the moving robot. Both intrinsic and extrinsic parameters are subject to change during operation. For instance, temperature changes might modify the imaging conditions of a camera lens, or the addition of a load weight might modify the pressure inflation on the robot tires, thus changing its kinematic parameters. An unforeseen collision of the robot in the workplace might modify the orientation of a sensor, changing its extrinsic calibration. The methods studied in this project will provide solutions to these problems, hence adding robustness to the company’s products.
