Trajectory planning for surface following with a manipulator under RGB-D visual guidance

Abstract

This paper introduces a manipulator robot surface following algorithm using a 3D model of vehicle body panels acquired by a network of rapid but low resolution RGB-D sensors. The main objective of this work is to scan and dynamically explore regions of interest over an automotive vehicle body under visual guidance by closely following the surface curves and maintaining close proximity to the object. The work is motivated by applications in automated vehicles inspection and screening in security applications. The proposed path planning strategy is developed based on a perceptionmodeling-planning-action approach. Raw data rapidly captured by a calibrated network of Kinect sensors are processed to provide the required 3D surface shape of objects, normal measurements, orientation estimation, and obstacle detection. A robust motion planning method is designed that relies on this information, resulting in a safe trajectory that is planned to follow and explore the curved surfaces while avoiding collision with protruding components of the vehicle. The feasibility and effectiveness of the proposed method is validated through experimental results with a 7-DOF manipulator navigating over automotive body panels.

Surface Following With An Rgb-D Vision-Guided Robotic System For Automated And Rapid Vehicle Inspection

“…In order to follow the vehicle surface within a given region of interest, the trajectory of the end effector is planned using a raster scan motion according to the 3D model defined in the previous section. The global path planning strategy, introduced in [24], assumes that each zone of interest is bounded in a rectangular box delimited by a set of points at the edge of the surface (Ymin, Ymax, Zmin, Zmax), corresponding respectively to the minimum and maximum values of the Y and Z coordinates in the vertex list (Fig. 10a).…”

“…As part of these initiatives, this paper introduces a fully automated vision-guided global path planning method that is adapted to explore and scan regions of interest over a vehicle body using a rapidly constructed 3D model of the vehicle. As such, this paper represents a technically extended version of our previous paper [24] on the complete development and integration of the sensing stage and the path generation process to handle more complex surface shapes.…”

Robot manipulator self-identification for surrounding obstacle detection

“…In [12], Natarajan uses a 3D Time-Of-Flight (TOF) camera as a proximity sensor to identify the position of the obstacle near the elbow of the manipulator. In [11], Danial introduces a manipulator robot surface following algorithm using a 3D model of the vehicle body panels acquired by a network of rapid but low resolution RGB-D sensors. The advantages of eye-in-hand arrangement are providing higher resolution point cloud of the obstacle, and that the algorithm is relatively simple because the coordinate transformation is not needed to measure the distance between the robot and the obstacle.…”

Safe Close-Proximity and Physical Human-Robot Interaction Using Industrial Robots

“…The user controls the robot motion remotely using an interface. The remote interaction approaches [7] [8] have been mainly proposed to direct the robots from a distance with the purpose of supporting and cooperating with humans in order to accomplish a specific task. In proximity interaction, the robot and the user work in close proximity to each other and they share the same space but typically there is no direct contact between the two [9].…”