Future Industrial Robotics - Springer978-3-319-03838-4/1/1.pdf · Future Industrial Robotics 26 ... experiment TRAFCON models from the AGV traffic control problem are used to . ...

Part I

Future Industrial Robotics

Future Industrial Robotics

26

Part I, Future Industrial Robotics, summarizes the results of five experiments funded by ECHORD focused on the challenges of industrial robotics. Industrial robots represent nowadays the vast majority of the operational robots worldwide but their use is still dominated by some traditional applications, like welding, assembly or handling. Furthermore, the majority of these installations still present low levels of autonomy/ flexibility, relying heavily on programming tasks, with reduced human-robot interaction during operation.

The human robot co-worker scenario is one of the key enablers for future robotics applications. The removal of fences or hard safety measures will in the future allow robots to perform more complex relying on close cooperation with humans. One of the crucial challenges for the robot co-worker scenario is the development of sensing systems that permit close and safe human robot interaction. In this scope the experiment EXECELL describes a novel projection-based sensor system that is intrinsically safe, can dynamically adapt the safety spaces and provides soft safety features such as the visibility of the space boundaries and visualization of intended robot movements.

The time required for robot installation and (re)programming represents a major technical and economical barrier for a broader use of industrial robots. With respect to industrial manipulators, programming is currently almost exclusively made via teach pendant which is not only time consuming but also economically not viable for many applications, since it is performed by trained technicians and requires the work cell to be stopped during programming. With respect to mobile logistics robots, or Autonomous Guided Vehicles AGV’s, installation and programming is still performed by engineering experts in a case by case procedure, limiting in this way even more the number of companies that can benefit from it. Therefore, some of the most significant research challenges for industrial robotics are related with increased levels of autonomy/flexibility and advanced human robot interfaces and programming. In the following part, two chapters are focused on the development of programming by demonstration techniques. In the experiments FREE and dimROB, In-Situ Robotic Fabrication: Advanced Digital Manufacturing Beyond the Laboratory, the use of human movement capture systems allows the robot to be programmed explicitly. In this scenario, the operator uses a tool to describe the trajectory that is afterwards automatically processed to produce valid robot programs. In the experiment MoFTaG, Kinesthetic teaching using assisted gravity compensation for model-free trajectory generation in confined spaces the operator can perform kinesthetic programming, i.e. through physical guidance of a robot with active compliance enabled by impedance-based control.

The needs for frequent reprogramming and reengineering of robotics systems call for increased levels of autonomy. The use of techniques such as machine learning, automatic planning or model driven programming brings industrial robotics autonomy, and consequently applicability, to a new level. In the experiment MoFTaG, the use of an extreme learning machine (neural network) and implicit scene modeling allows the transfer of the user’s implicit knowledge via kinesthetic programming. In the experiment TRAFCON models from the AGV traffic control problem are used to

Future Industrial Robotics

27

design a novel traffic manager that is more efficient and flexible for different industrial setups and above all can dynamically change the paths robots are following.

Finally, the Future of Industrial Robotics also builds on top of new applications. Among the areas where robotics presence is weak or null the construction industry case is one of the most outstanding due to its economic size and importance. The experiment dimROB comes from a robotic laboratory that has a long tradition in digital manufacturing. In this experiment, the use of a mobile manipulator opens great perspectives in terms of large scale digitally generated architectural pieces.

Below, a brief synopsis of each experiment in Part I is presented. The experiment EXECELL presents the application of a novel projection-based

safety system capable of ensuring hard safety in human-robot collaboration. It presents the enhancement of the development of newly developed projection-based safety system to utilize the current state of the robot to obtain optimizes safety spaces and describes the machine vision algorithms required for robust violation detection. Furthermore presents an in depth discussion about the compliance of the system with the IEC standards

The experiment FREE presents a flexible and safe interactive human-robot environment, achievable through a combination of standard commercial robot with the state of the art safety and control technologies. The experiment presents a control loop, the Superior Hierarchical control that interfaces the human and the robot with human position detection and operator work recording for task learning. The efficient robot learning procedure allows reducing or eliminating the need of accurate positioning of the workpiece with jigs and fixtures.

The experiment dimRob, In-Situ Robotic Fabrication: Advanced Digital Manufacturing Beyond the Laboratory fosters a non-standard digital robotic fabrication using a mobile manipulator that can be directly applied on the construction site and it is easily scalable. The experiment deals with challenges in robotics, like tolerance handling and human robot interaction but also with the architectural implications of the integration of the findings that resulted from experimentation at the earliest stages of design. The results show mobile manipulator that is capable of performing on situ digital fabrication of complex structures using advanced sensing for tolerance handling, perform self-localization and also being programmed explicitly by the user.

The experiment TRAFCON proposes a novel traffic manager that capable of efficiently control the coordinated motion of the AGVs and dynamically adapt the paths the robots are following. TRAFCON traffic manager was experimentally validated on simulated real plants and on a small-scale automatic warehouse and shown the significant reductions in the completion time through the introduction of dynamic routing.

The experiment MoFTaG, Kinesthetic teaching using assisted gravity compensation for model-free trajectory generation in confined spaces, presents approaches programming of redundant robot manipulator in a co-worker scenario from a user-centered point of view. In contrast with other kinesthetic approaches, in the MoFTag experiment the operator rather configures the robot by providing training data in different areas of the robot’s workspace for the learning algorithm to infer an appropriate redundancy resolution. In a subsequent stage, the system provides assistance to the operator about a learned or configured redundancy solution.