We present a learning-based method for 6 DoF pose estimation of rigid objects in point cloud data. Many recent learning-based approaches use primarily RGB information for detecting objects, in some cases with an added refinement step using depth data. Our method consumes unordered point sets with/without RGB information, from initial detection to the final transformation estimation stage. This allows us to achieve accurate pose estimates, in some cases surpassing state of the art methods trained on the same data.
To support shifting to high mix/low volume production, manufacturers in high wage countries aim for robotizing their production operations – with a special focus on the late production phases, where robotic assembly cells are then confronted with any complexities resulting from part and product varieties. The ‘World Robot Challenge 2018’ (WRC 2018) emulated such high mix/low volume production scenarios in a competition taking place in Tokyo, Japan. As part of our activities in SDU's newly founded I4.0 Lab, we integrated and advanced our experiences and developments from our various R & D projects in a novel robotic assembly cell design to compete in the WRC 2018. This article describes the system architecture as well as main aspects of its implementation regarding robot control, robot programming and computer vision and how they contributed to winning the challenge. Due to the application of collaborative robots, the cell design allows for operation without fences. Hence, multiple copies of the cell can be arranged in a highly reconfigurable, highly adaptable matrix structure in which several production flows can be handled concurrently. This concept was demonstrated by the installation of a duplicate cell that allowed for parallel developments on two cells and prolonged development also after shipping the first cell to Japan.
This paper presents a method for high precision visual pose estimation along with a simple setup procedure. Robotics for industrial solutions is a rapidly growing field and these robots require very precise position information to perform manipulations. This is usually accomplished using e.g. fixtures or feeders, both expensive hardware solutions. To enable fast changes in production, more flexible solutions are required, one possibility being visual pose estimation. Although many current pose estimation algorithms show increased performance in terms of recognition rates on public datasets, they do not focus on actual applications, neither in setup complexity or high accuracy during object localization. In contrast, our method focuses on solving a number of specific pose estimation problems in a seamless manner with a simple setup procedure. Our method relies on a number of workcell constraints and employs a novel method for automatically finding stable object poses. In addition, we use an active rendering method for refining the estimated object poses, giving a very fine localization, suitable for robotic manipulation. Experiments with current state-of-the-art 2D algorithms and our method show an average improvement from 9 mm to 0.95 mm uncertainty. The method was also used by the winning team at the 2018 World Robot Summit Assembly Challenge.
Vision systems are often used jointly with robotic manipulators to perform automated tasks in industrial applications. Still, the correct set up of such workcells is difficult and requires significant resources. One of the main challenges, when implementing such systems in industrial use cases, is the pose uncertainties presented by the vision system which have to be handled by grasping. In this paper, we present a framework for the design and analysis of optimal gripper finger designs and vision parameters. The proposed framework consists of two parallel methods which rely on vision and grasping simulation to provide an initial estimation of the uncertainty compensation capabilities of the designs. In case the compensation is not feasible with the initial design, an optimization process is introduced, to select the optimal pose estimation parameters and finger designs for the presented task. The proposed framework was evaluated in dynamic simulation and implemented in a real industrial use case.
The gripper finger design is a recurring problem in many robotic grasping platforms used in industry. The task of switching the gripper configuration to accommodate for a new batch of objects typically requires engineering expertise, and is a lengthy and costly iterative trial-and-error process. One of the open challenges is the need for the gripper to compensate for uncertainties inherent to the workcell, e.g. due to errors in calibration, inaccurate pose estimation from the vision system, or object deformation. In this paper, we present an analysis of gripper uncertainty compensating capabilities in a sample industrial object grasping scenario for a finger that was designed using an automated simulation-based geometry optimization method (Wolniakowski et al., 2013, 2015). We test the developed gripper with a set of grasps subjected to structured perturbation in a simulation environment and in the real-world setting. We provide a comparison of the data obtained by using both of these approaches. We argue that the strong correspondence observed in results validates the use of dynamic simulation for the gripper finger design and optimization.