Thumbnail Image

BIM semantic enrichment for the path planning of indoor wall-climbing robots

Leyuan , Ma; Timo, Hartmann (Contributor)

The availability of small-sized and compact wall-climbing robots is expected to facilitate inspection, maintenance, and domestic service tasks, especially in crowded indoor spaces where floor spaces are largely occupied. In contrast to the Unmanned Ground Vehicles (UGVs), the operation of wall-climbing robots would cost more power since it requires extra energy to overcome their gravity and adhere stably to inclined surfaces. Efficiently finding an optimal path between two different locations is crucial for the wall-climbing robot to improve energy efficiency and is the key to promoting the application of wall-climbing robots indoors. To navigate, the robot should construct or possess an appropriate representation of the surrounding environment. The simultaneous localization and mapping (SLAM) approach is a common way used by Unmanned Ground Vehicles (UGVs) to create a map of an unknown environment. To achieve this, the robot needs to be equipped with sensors (e.g.2D or 3D laser scanner) to scan its surroundings and an on-board processor to process the collected data in real-time. This process often requires a high onboard processing capacity and costs a lot of energy, which is challenging for indoor wall-climbing robots due to the limited payload capacity and battery power. Furthermore, traditional SLAM approaches only output geometric maps without semantic features, which might result in inaccurate or unreasonable path planning in practical applications. Most research on semantic mapping in the robotics field relies on RGB-D cameras equipped on the robot and computer vision algorithms for semantic detections. This, however, causes an increase on the computational cost. In the AEC industry, Building Information Modelling (BIM), supported by the international standard Industrial Foundation Class (IFC), is the most well-known approach to represent comprehensive geometric and semantic information of a building. The widespread adoption of BIM and the availability of increasingly reliable BIM models offer the wall-climbing robot new sources of information about their surroundings. Instead of solely relying on on-board sensors, making use of BIM data would considerably reduce the computational burden of the robot. Previous studies on BIM-based indoor path planning have proposed different indoor data models suitable for the path planning of different users including pedestrians, emergency responders and ground mobile robots. The users considered in these studies usually move in a horizontal direction. However, wall-climbing robots have different movement patterns and route preferences. The information included in existing data models cannot meet the requirement of wall-climbing robot path planning.Therefore, to facilitate BIM-based path planning of wall-climbing robots, a new indoor data model needs to be developed and the voxelization algorithm needs to be improved.