The technical and technological advances that have been achieved over the past decades have led to a tremendous increase of both the types as well as the total amount of electrical and electronic equipment that is manufactured by the industry. On the other hand, the lowering of the industrial production cost together with the continuous and rapid change in the technology have resulted in the wide spread use of the produced devices in large quantities, along with the continuous need to often being upgraded/replaced. The above facts have led to the generation of enormous amounts of Waste Electrical and Electronic Equipment (WEEE). The importance of managing the WEEE materials and its tremendous impact on the economy, society and environment is easy to be realized, by considering that the production of common electrical and electronic equipment (e.g. smartphones, PCs, monitors, tablets, home appliances, etc.) requires highly expensive and often rare material types (e.g. gold, copper, steel, etc.). Despite the great importance and the tremendous economic/environmental/societal impact of WEEE management, the current technology are still confronted with challenges, such as the hazardous WEEE materials, human and environmental risks, illegal activities, recycling cost and strict policies. Therefore, a ‘hybrid human-robot recycling plant for electrical and electronic equipment’ operating in an indoor environment for the recycling of WEEE is desired.
The ultimate goal of HR-Recycler is to develop a sophisticated open human-robot working environment that will implement a ‘hybrid recycling plant for electrical and electronic equipment’, operating in an indoor setting. The fundamental principle behind the design of the envisaged system is the replacement of multiple currently manual, expensive, hazardous and time-consuming tasks of WEEE materials pre-processing with correspondingly automatic robotic-based procedures, fused within a genuine human-robot collaboration context that will boost the productivity and quality of work in the plant. The primary output of the envisaged system will be to extract sorted electric/electronic device components and concentrated fractions of increased economic and environmental value; hence, contributing to the fundamental goal of the ‘European circular economy’ project and boosting economic activity in secondary markets.
The HR-Recycler system has been designed based on real user needs, while for its eventual evaluation pilot studies in real-world operational environments are foreseen. The human could collaborate, by improving the equipment placement as well as picking up unknown equipment. To demonstrate the capabilities of the proposed system, four specific types of WEEE have been selected, which will constitute the following four main Use Cases.
Emergency lamps are one of the most common WEEE materials that originate in principle from the construction sector, i.e. when renovating or demolishing old buildings. The main goal during the manual disassembly process is to remove the fluorescent lamps without causing any damage to it, since they are potentially hazardous elements to the human health and the environment.
Microwave ovens constitute the type of domestic appliances with the most increased recycling rate, compared to other types of devices, mainly due to their smaller lifecycle, greater frequency of malfunction and lower repair capabilities. In terms of recycling needs, ovens present numerous potential cases (e.g. motors, glass and plastic components, metal integument, etc.).
The very large expansion of the use of Personal Computers (PCs) and their continuous need for upgrade results in a corresponding tremendous generation of WEEE material of this type. The critical characteristic about PC towers is that they include components that can be re-used to a great extent (e.g. motors, CPUs, batteries, plastic cases, etc.), while they also include significant quantities of rear materials (precious metals, high performance plastics, etc.).
Both Flat Panel Displays (FPDs) and CRT monitors are considered. FPD is one of the most commonly met WEEE material types and it is generated in extremely high quantities. One of the fundamental goals of the respective recycling pipeline is the initial classification in LCD and LED monitor types. Regarding CRTs, although they are no longer commercialized, the number of screens reaching treatment plants is still enormous.
In the HR-recycler project, LSR is responsible for the development of novel algorithms and control methods for force guided manipulation, flexible probing, versatile grasping and human-aware motion planning to ensure safe robot manipulation. Holding the advantages of affluent research expertise and academic resources, LSR also focuses on highly ranked venues for publishing research results, including academic journals, national and international conferences and workshops, as well as more specialist events related to the core research areas of HR-Recycler: human-robot interaction, human action modelling and prediction. LSR is offering a prime opportunity for students to learn the state of the art technology of human-robot collaboration in real-world industrial settings and improve their chance for a job in industry or in research, especially the topics for research theses at Bachelor/Master/PhD levels to advance further on the success of HR-Recycler.
A fundamental challenge in disassembling unknown objects are manipulation tasks with uncertain exact object dimensions and locations, e.g. the location of screws or the outer perimeter of objects are captured through camera systems, involving measurement uncertainty. This uncertainty can be accounted for by controlling the lateral force applied to the tool, thus, allowing the tool so “snap” into the desired location. However, the envisioned disassembly scenario requires different control strategies along the task axes and varying impedance of the controlled force depending on the structure of the object to be disassembled. In this project, we will devise a unified control policy capable of regulating both the robot motion and its physical interaction with the environment. This control policy is learned from human demonstrations. The robot motion and its stiffness behaviours are modelled including the desired damping throughout the motion. Therefore, the approach is suitable for generating motions that follows the same velocity profile as found in human demonstrations. The controller is a unification approach between “real-time motion generation” and “variable impedance control”, with the advantage that it guarantees stability as well as does not rely on following a reference trajectory.
Concerned Researcher: Volker Gabler
The shell of modern devices is usually held together through multiple fixtures. Such fixtures can be reversible, such as screws, or non-reversible, such as plastic snap-in fixtures. When disassembling devices without knowing their construction plan, fixtures are easily missed – either because they are not visible, such as snap-in fixtures, or because they are not properly recognized, such as recessed screws. Therefore, we proposed an approach, where a robot opens all recognized fixtures by unscrewing all screws and cutting open the perimeter of the enclosure. In a second step, we mimic the behaviour of humans by prying the device case along the cut opening and flexing the cut-off case. We seek to identify the location of missing fixtures through probing the case structure by bending it from multiple locations and identify a model for the structure and possible remaining fixtures. We propose an approach, where haptic and visual modality are jointly used to learn and identify object structures and construction properties, which integrates control structures with learning algorithms, to identify physical probing locations to obtain a maximum reduction of model uncertainty with each probe.
Some components that have to be removed, need to be handled with care e.g. mercury lamps in LCD flat screens, fluorescent lamps in emergency lamps etc. This requires force-adaptive grasping strategy with tactile sensing gripper fingers. At all cost it must be avoided that forces and torques are applied in directions other than the ones desired for disassembly. This can only be achieved if sensitive and reliable measurements are available. Thus, the first task is to evaluate suitable hardware components, and develop a sophisticated sensor fusion method. Besides force-adaptive task-optimal gripper alignment, the gripper grasping force is the second crucial component in successful disassembly. For the generalisation of robotic grippers, the implementation of a coupling of feedback and feedforward controllers is proposed that aim at adjusting the gains of the forces applied based on sensory feedback and predictions. The goal of the sensory predictions is to minimise a given error by acting in anticipation. The proposed controller is based on the cerebellum and it can result in successful adaptive grasping, without having to implement force controllers for each object.
In human-machine collaboration, the robot not only needs to satisfy physical constraints due to the limit on mechanical systems but also has to guarantee safety for itself and its human collaborators. Therefore, the robot has to evaluate available actions to be taken, interpret human motion with respect to possible tasks currently executed by the human, and select and execute its own action accordingly. Such a safe interaction also requires planning legible motions for the robot that can easily be understood by its human partner. In that regard, one of the key aims of this task is to develop novel stochastic trajectory optimization and constrained policy improvement methods to enable the robot to proactively collaborate with human partners. The safety-control model will address the following issues: robot safety (by ensuring that the performed actions do not harm the robot), material safety and human physical and perceived safety, which includes the modulation of the speed and amplitude of the robot’s movements, the speed of the conveyor belt and the distance of the robot from the human co-worker. Regarding navigation, the model will include obstacle and human avoidance as well as modulation of the speed during the transportation of the material. The outcomes of this task will contribute generation of the robot’s actions.
Zhang, Z.; Qian, K; Schuller, B. W.; Wollherr, D.: An Online Robot Collision Detection and Identification Scheme by Supervised Learning and Bayesian Decision Theory. IEEE Transactions on Automation Science and Engineering, 2020, 1-13 [mehr…] [ Volltext ( DOI )] [ Volltext (mediaTUM)]
Hoang Dinh, Khoi.; Oguz, Ozgur. Salih.; Elsayed, Mariam.; Wollherr, Dirk.: Adaptation and Transfer of Robot Motion Policies for Close Proximity Human-Robot Interaction. Frontiers in Robotics and AI, 2019, 6, 69. [Full text (DOI)]