Weston Robot built a fully autonomous inspection robot with school of EEE at NTU, Singapore, to enable researchers to develop AI-based applications for construction inspection. The entire robot is ROS compatible with extensive interfaces for further development. With this autonomous robot, researchers from NTU can quickly leverage to test, validate, and deploy their research.
The robot consists of three parts: a 4-wheel differential drive robot base, a 3D-lidar based navigation stack, and vision-based inspection unit. Considering rough terrains at construction sites, researchers from NTU chose SCOUT as the robot base. The robot has zero turning radius, is also capable of carrying 50kg payload, climbing 20-degree ramp and passing through 15cm obstacles.
For outdoor navigation, 3D-lidar based navigation stack is often adopted. Engineers team from Weston Robot developed a 3D-lidar based navigation to enable the robot to move autonomously in unconstructed environment. Besides 3D lidar data, RTK (differential GPS), RGBD, and Ultrasonic sensors can also be fused to enhance the 3D navigation capabilities, e.g., RGBD and Ultrasonic sensors are also used to do collision avoidance in the dead zone of the 3D lidar scanning range.
To enable vision-based AI application, e.g., detecting abnormalities at construction sites, edge-based computational and streaming unit is recruited. The robot equips 360° HD cameras, which consists of 6 camera lens pointing to different directions, and can do live streaming to the edge computer, or to the remote data centre. Besides the 360° HD cameras, an FHD PTZ camera is also installed to capture the detailed images/videos for further analysis.