The `following game` - Towards interactions in groups of heterogeneous robots
Kerstin Dautenhahn (VUB AI-Lab, Belgium, firstname.lastname@example.org)
Description of the demonstration
(see also photos of the demonstration in Nara. )
The demonstration shows an `interactive experiment` with a group of heterogeneous autonomous robots. The scientific background of the demonstration is the investigation of `social interactions` in groups of heterogenous robots. `Social` refers to direct interactions between individuals. The level of abstraction to be investigated is therefore the individual with its characteristic sensory-motor capabilities and its unique way of interacting with the social and nonsocial environment. One central behavior which we study in such `individualized` groups of robots is keeping-contact and following behavior. `Guiding` can be used in groups of heterogenous robots as an indirect (without explicit communication) method of `situated teaching`, which means that information about behavior and context is not transmitted but `re-experienced` by the following agent.
Snapshots of the following game:
The robots which are shown in the demonstration have originally been designed for specific `habitats`, either a hilly landscape (the `Huegellandschaft`, developed at GMD, Germany) or an ecosystem for self-sufficiency experiments (VUB). The demonstration shows a simple scenario with interactions between these special-design robots. The robots morphologies and sensor equipments have not been specifically adapted to the purpose of the demonstration.
Five different kinds of agents can take part in the `following game`: the role of the guiding agent is either taken by a human or a remote controlled vehicle (controlled itself by a human). So there are two possible ways in which humans can interact with the robots in order to play a `guiding role` or to test the dynamics of the robots` behavior. The followers could be three different robot `species`: two ftbots (constructed with fischertechnik components), one with rectangular shape (thanks to Guenter Frick) and one with oval shape, plus one Cbot (Lego based robot, built by Peter Stuer at VUB). The followers could also form chains and consequently take the role of a `guiding agent`.
The three robot `species` show different strategies to implement keeping-contact and following behavior. After forming the global pattern of a chain the strategies can hardly be distinguished, so that from a observer point of view the complexity of the control programs is then not reflected by its performance. But the differences become visible by interactively `testing` the robustness of the control strategies. The different control strategies for keeping contact and following are:
- aligning towards any object in the robot`s preferred direction of movement (using IR sensors)
- alignment of the robot`s body axis in order to establish `bodily` contact to an object at a desired `contact zone` (binary contact sensors)
- `actively` following a moving object, distinguishing and testing moving and nonmoving objects (using IR sensors and contact sensors)
Acknowledgement: technical support was given by GMD, Research Division Artificial Intelligence, Germany
Back to Kerstin`s Home Page.
1996-05-01 Kerstin Dautenhahn