https://members.femto-st.fr/julien-bourgeois
https://www.programmable-matter.com
Technological advances, especially in the miniaturization of robotic devices foreshadow the emergence of large-scale ensembles of small-size resource-constrained robots that distributively cooperate to achieve complex tasks. These ensembles are formed by independent, intelligent, and communicating units which act as a whole forming a programmable material i.e., a material able to autonomously change its shape.
In my talk, I will present our research effort in building a modular robot composed of mm-scale units. We use micro-technology to scale down the size of each element, and we study geometry, structure, actuation, power, electronics, and integration. We develop multi-agent algorithms to scale up in the number of managed robots to perform synchronization, leader election, self-assembly and self-reconfiguration. As multi-agent systems are by essence decentralized, they are the best candidate to manage these distributed robotic systems.
Large language models have stretched our beliefs on what kind of machine intelligence is possible. However, there are still practical needs in several fields of computing to align artificial intelligence (AI) with human intelligence.
In this talk, I will discuss and problematize:
I will give insights on how cognitive mimetics can inspire AI solutions and review on-going research on modeling human intelligence as computational rationality with reinforcement learning agents in our cognitive science research group at University of Jyväskylä.