No abstract

Conventional feedback control methods can solve various types of robot control problems very efficiently by capturing the structure with explicit models, such as rigid body equations of motion. However, many control problems in modern manufacturing deal with contacts and friction, which are difficult to capture with first-order physical modeling. Hence, applying control design methodologies to these kinds of problems often results in brittle and inaccurate controllers, which have to be manually tuned for deployment. Reinforcement learning (RL) methods have been demonstrated to be capable of learning continuous robot controllers from interactions with the environment, even for problems that include friction and contacts. In this paper, we study how we can solve difficult control problems in the real world by decomposing them into a part that is solved efficiently by conventional feedback control methods, and the residual which is solved with RL. The final control policy is a superposition of both control signals. We demonstrate our approach by training an agent to successfully perform a real-world block assembly task involving contacts and unstable objects.

No abstract

Virtual assembly training systems show a high potential to complement or even replace physical setups for training of assembly processes in and beyond the automotive industry. The precondition for the breakthrough of virtual training is that it overcomes the problems of former approaches. The paper presents the design approach taken during the development of a game-based, virtual training system for procedural assembly knowledge in the EU-FP7 project VISTRA. One key challenge to address when developing virtual assembly training is the extensive authoring effort for setting up virtual environments. Although knowledge from the product and manufacturing design is available and could be used for virtual training, a concept for integration of this data is still missing. This paper presents the design of a platform which transfers available enterprise data into a unified model for virtual training and thus enables virtual training of workers at the assembly line before the physical prototypes exist. The data requirements and constraints stemming from industrial partners involved in the project will be discussed. A second hurdle for virtual training is the insufficient user integration and acceptance. In this context, the paper introduces an innovative hardware set-up for game-based user interaction, which has been chosen to enhance user involvement and acceptance of virtual training

No abstract