ABOUT    |    MEDIA    |    CONTACT    |    CAREERS
  Precision Tunnel Mapping

Example imagery from the sensor in an underground mine and a rendered view of a model.

Current State of Tunnel Mapping Technology
State-of-the-art devices for tunnel and subterranean void mapping generally fall into one of two categories: statically positioned single point lidars, and vehicle mounted planar lidars. Mapping of tunnels requires that successive scans have high point cloud overlap to ensure proper registration and stitching. Most tunnels do not allow for a long line-of-sight, requiring scans to be taken at frequent spatial intervals leading to very slow mapping speeds. Actuated planar lidar systems mounted on vehicles are much faster but face the challenge of needing to estimate vehicle motion during lidar scanning. Traditionally, the motion is estimated via an IMU. However, this approach is limited as lower power, small form factor IMU’s have significant drift rates and are expensive. They also often have export control restrictions.

Our approach combines state-of-the-art visual odometry with lidar scanning to both provide the estimate of vehicle motion and the initial solution for pose optimization. The system runs in real time, on limited computing. 30W powers all sensors, computing, lighting, and actuation. The whole system is small enough to run on a snake robot and is little larger than a coke can.

Experiments in coal mining operations have demonstrated the robustness and accuracy of the system. Moreover, we have developed highly accurate calibration systems to ensure that the open loop accuracy is high. This is crucial as many tunnel environments (e.g. smuggling tunnels) may not have opportunities for “loop closures” to reduce overall error build up.

National Robotics Engineering Center    SITE MAP    |    ABOUT    |    MEDIA    |    CONTACT    |    CAREERS    © Carnegie Mellon University