KUKA Innovation Award 2016

The CoSTAR team establishes the basis for enabling end users to easily specify complex task models. This will support users in a flexible manufacturing environment by incorporating object information – and allowing them to react to external events.

The goal is to break away from programming robots with linear plans for specific tasks, and instead to teach robots on the basis of libraries of skills and known objects. In this way, users can create complex, adaptive task plans to fit new situations.

The goal of this demonstration project is to develop a safe and intuitive human-robot collaboration application in the context of the automotive industry. In this application, a robot has to locate and manipulate similar, but not identical workpieces using force, 3D image processing and the fusion of laser sensors. The robot has to pick up workpieces, place them in plastic trays, pushing them into position and bending the sides of the tray over if necessary.

The operator can interact with the tray and the products. This changes the situation during the ongoing process and the robot has to adapt constantly to the new situation. It senses and interprets its environment using artificial intelligence and acts accordingly. This enables the robot to interact safely with the operator.

The “Bin Picking Assistant” application combines the potential of human-robot collaboration with a reading system. In doing so, the application creates a flexible work assistant.

In the project for the KUKA Innovation Award, an existing reading software component is interfaced to the LBR iiwa. This software enables the robot to pick randomly positioned workpieces from a bin and hand them to the operator. The operator can line this part up with another workpiece – for example in a screw fastening task. The LBR iiwa holds the workpiece during this process, thereby compensating for gravity. It simultaneously has to master the challenge of ensuring a stable reading process. An extremely important aspect here is human-robot collaboration. For this purpose, the collision detection parameters of the LBR have to be correctly set.

To enable robotics to be used in new fields with novel applications, the robot programs must be adapted to their environment. Due to their complex nature, construction sites may be regarded as an ideal testing environment for the universal deployability of robots. Because of the fluctuation and limited experience of personnel, it must be possible for a robot to be used, calibrated and reprogrammed for construction tasks by users with no technical background.

The team therefore proposes a new approach to robot programming. By combining research results from the field of teaching by demonstration with skill and parameter-based robot programming, the team aims to redefine the current teaching process. The “Dynamic and Interactive Robotic Assistant for Novel Applications” (DIANA) is intended to demonstrate the opportunities offered by this approach in the assembly of building elements for the construction site of the future.

The project deals with the development of a robust assembly operation integrating the capabilities of the KUKA flexFELLOW and an image processing system. The developed system will be capable of adapting to a new environment and performing an assembly task for multiple workstations using an onboard setup.

The main contributions of the project are a routine for robot calibration using kinesthetic teaching, product and workspace recognition using machine vision, collision-free robot trajectory generation and an autonomous assembly operation using the force sensing, Cartesian stiffness and forced oscillation capabilities of the KUKA LBR iiwa robot. The project aims to reduce the time required for robot programming in semi-structured environments for assembly operations.

With the “WEIR” project, which stands for Wearable Interaction with Robots, the team aims to overcome the current limits of robotic co-working. To achieve this, factory workers are to be furnished with the capability of teaching robot arms new poses and work steps with the aid of wearables. This enables easy, fast, cost-effective and reliable teaching of tasks for robot arms, personalized control of the robot arm and adaptation of the control software runtime based on dynamic properties of the environment.