Edouard Ramstein

Robot mechanism synthesis and genetic algorithms

We propose to explore the problem of mechanism synthesis with an approach using genetic algorithms. We intend to study the case where there exist geometric constraints (as avoiding obstacles). More precisely, we are interested in the choice of robots (morphology and position) which has to follow a given trajectory among obstacles.

Emergence of a Virtual Human Behaviour in a VR Environment

Nowadays, time to market becoming shorter and shorter, physical mock up is being more and more often replaced by digital mock up. However physical mock up was very useful to the present day to check accessibility for example. So, it is now important to have tools available to simulate with digital mock up what physical mock up used to be used for. For example, to check accessibility, a tool that allows the simulation of the manipulation of an object by a manikin in a cluttered environment without collision is needed. In this paper, we propose a new way to create human behaviour of a manikin using a multi-agent approach. Our system is based on a former architecture already presented. It consists in a collision-free path planning system using cooperation between software and human operator abilities. The architecture for this co-operation is based on a multi-level tree architecture, which enables the architecture to be completed by adding as many agents as we need. The conflicts between agents are managed by acting on their period of activity. We demonstrate in this paper the fact that collaboration between many agents, each providing the manikin with an elementary behaviour, can provide the manikin with a coherent global virtual human behaviour.