SWARMIX - Synergistic Interactions in Swarms of Heterogenous Agents
The proposed research is about laying the foundations for the design, implementation, and adaptive control of heterogenous multi-agent systems that are composed of humans, animals, and robots working in cooperation to solve distributed tasks that require a wide diversity of sensory-motor and cognitive skills. In the following, we will refer to such systems as mixed swarms, hence the project code name SWARMIX. The aim is to provide each component of the mixed swarm with a high level of autonomy in order to allow it to fully exploit its own unique skills and abilities, and at the same time to set up close bidirectional interactions and information flows between all system components in order to ensure overall synergistic cooperation. The main novelty of this project lies in the cooperative integration of a possibly large number of humans, animals, and robots in tight cooperation in one single networked system with distributed control. To the best of our knowledge this would be the first research addressing peer-to-peer cooperation between humans, animals, and robots. The research work in the SWARMIX project will be guided by the creation of a specific validation scenario composed of humans, dogs, and small-sized UAVs aimed at performing search-and-rescue (SAR) missions in outdoor civilian domains. We believe that concentrating on one example is important given the vast possibilities in the ways mixed swarms can be composed and applied. A practical implementation will help guide the development of theoretical work and provide a testbed for new ideas. The choice for search-and-rescue missions is inspired by their practical/humanitarian relevance, and by the fact that these applications are often used in the scientific domain as reference test to assess the performance of collective cooperative systems. In order to validate our research and to emphasize the synergistic cooperation between partners, the SWARMIX project will include a rigorous experimental protocol in the context of the SAR scenario, where success will be measured by the capability of the swarm as a unit to locate targets in distributed environments more reliably and/or faster than each group of agents in isolation or even in pairwise combinations. Our research will focus on the one hand on the identification and development of technology that is needed to use and extend the sensory-motor skills of the different agents and to support communication between them, and on the other hand on the creation of intelligent algorithms and protocols that allow synergistic interaction and control within the heterogenous swarm. To break down the complexity of the general problem at hand, we identify four fundamental directions of research, that are addressed in the individual sub-projects: 1) wireless networking for highly dynamic heterogenous agents studied at ETHZ-TIK, 2) adaptive profiling of agent and mission status, and swarm-level planning of agent activities studied at IDSIA, 3) intuitive inter-agent communication interfaces and local control of agent behaviors studied at EPFL-LIS, 4) training, control, and monitoring of dogs for full integration in the swarm studied at ELU. Each of these four research directions stands by itself and addresses fundamental scientific issues. It is from the interaction among these four research directions and their integration into one common view and real-life application that we can obtain a true synergy of results and provide answers to the main challenge of this project, namely adaptive cooperation in heterogenous, multi-agent systems to perform tasks more efficiently.