The ERICA project tries to further the development of human-robot-cooperation systems in the industrial domain. The project comprises the construction and the evaluation of a sophisticated cognitive system, which provides the safety of the human worker without the necessity of artificial boundaries like safety fences.
To meet the safety criteria, real-time 2D and 3D machine vision is deployed to provide estimates about objects in the scene, for instance the pose and position of the human worker. Based on this information, an elaborate cognitive layer compiles semantic data about activities and events. Together with a proactive task and path planner, this forms a system which will render safe human-robot-cooperation feasible. The evaluation will cover both, the evaluation of individual system modules and the evaluation of the overall system performance during the cooperation process. In addition to that, the feasibility and practicability of human-robot-cooperation itself will also be investigated.
For the evaluation process, different instances of common task classes in the industrial domain will be carried out in cooperation. Although quantitative measures like distance and time seem appropriate for the system evaluation, for the practicability review of the human-robot-cooperation, qualitative measures appear more meaningful. Therefore, a group of subjects will perform certain tasks in cooperation with the robot, and afterwards complete a questionnaire in order to provide qualitative information about the cooperation procedure. It should be mentioned that the sensor data used for the system evaluation will be real sensor readings as well as synthetic data produced by a rendering pipeline. The advantage of the latter is that synthetic data is immediately ground truth data. It is also easy to build distinctive test cases which instantiate certain events or actions.
The ERICA project is a part of the ECHORD framework, which allows for collaboration with our partners Reis Robotics and KUKA, merging scientific and commercial interests to gain results in both areas. The experimental results are considered to be relevant for the research in the field of human-robot-cooperation and the design of advanced cooperative systems.