This work presents a concept for autonomous mobile manipulation in industrial environments. Utilizing autonomy enables an unskilled human worker to easily configure complex robotics system setup phase before carrying out fetch and carry operations the execution phase. In order perform given tasks real production sites, we propose robotic consisting of platform, torque-controlled manipulator, additional sensor head. Multiple sensors are attached which allow perception environment objects be...

We present a next-best-scan (NBS) planning approach for autonomous 3D modeling. The system successively completes model from complex shaped objects by iteratively selecting NBS based on previously acquired data. For this purpose, new range data is accumulated in-the-loop into surface (streaming reconstruction) and continuous scan paths along the estimated trend are generated. Further, space around object explored using probabilistic exploration that considers sensor uncertainty. This allows...

The procedure of manually generating a 3D model an object is very time consuming for human operator. Next best-view (NBV) planning important aspect automation this in robotic environment. We propose surface-based NBV approach, which creates triangle surface from real-time data stream and determines viewpoints similar to intuition. Thereby, the boundaries are detected quadratic patch each boundary estimated. Then several viewpoint candidates calculated, look perpendicular overlap with...

Active scene exploration incorporates object recognition methods for analyzing a of partially known objects and approaches autonomous modeling unknown parts. In this work, recognition, exploration, planning are extended combined in single system, enabling advanced techniques such as multi-view from planned view positions iterative by integration new scene. Here, geometry based approach is used i.e. matching database. Unknown autonomously modeled added to the Next-Best-View performed both...

A robotic system for search and rescue missions needs to efficiently explore map new areas. In this paper, we present an integrated exploration strategy with active loop closing, which balances between the speed quality. Specifically, it finds a trade-off moving towards unknown space gather information revisiting previous locations improve localization accuracy quality through closures. Our is built upon submap-based 6D SLAM system. Loop closure constraints originate from pairwise submap...

Pick and place applications are very common in industrial environments. Autonomous execution of this task is desirable, as it simple repetitive for a human. To end, we employ mobile robot containing light weight (LWR) arm, pan-tilt unit (PTU) multitude sensors (see Fig. 1).

This paper discusses the potential benefits of plenoptic cameras for robot vision during on-orbit servicing missions. Robot is essential accurate and reliable positioning a robotic arm with millimeter accuracy tasks such as grasping, inspection or repair that are performed in close range to client satellite. Our discussion camera technology provides an overview conceptional advantages regard conditions mission. A camera, also known light field basically conventional system equipped...

This paper describes a multi-view pose estimation system, that is exploiting the mobility of depth sensor through mounting it onto robotic manipulator. Given algorithm performs feature extraction and matching to model database, we investigate probabilistic modeling space as well measurement uncertainty, be used in sequential state approach. Uncertainties 3d position can modeled parametric way by Gaussians, but rotations - special orthogonal group SO(3) requires approaches from directional...

As the performance of key perception tasks heavily depends on their parametrization, deploying versatile robots to different application domains will also require a way tune these changing scenarios by operators. many tunings are found trial and error basically experts as well, quality criteria change from application, we propose Pipeline Optimization Framework that helps overcoming lengthy setup times largely automating this process. When deployed, fine-tuning optimizations presented in...

This work focuses on Monte Carlo registration methods and their application with autonomous robots. A streaming an offline variant are developed, both based a particle filter. The is performed in real-time during data acquisition laser striper allowing for on-the-fly pose estimation. Thus, the acquired can be instantly utilized, example, object modeling or robot manipulation, scan aborted after convergence. Curvature features calculated online estimated poses optimized weighting step. For...

This work contributes the optimization of a streaming pose estimation particle filter and its integration into an autonomous object modeling approach. The is advanced by additional in weighting step. By integrating method approach, repositioning objects enabled, which often necessary order to acquire complete models. Experiments show that usage iterative closest point too restrictive for general transformations. used Monte Carlo enables robust without loss time with high precision. Further,...

This work is focused on streaming particle filter registration of surface models such as homogeneous triangle meshes and point clouds. Part the approach a curvature feature calculation. The investigated utilizes to incrementally update pose estimates during data acquisition. method evaluated in real experiments with high-precision laser striper system attached an industrial robot. During scan, integrated on-the-fly order calculate features based these estimate object's pose. Experiments show...