The two central goals of this project in computational engineering are:

  • to devise new strategies for the real-time distributed control of Complex Systems;
  • to develop a novel design methodology meant for Complex Systems.

These goals are considered for Complex Systems made of many identical interacting or communicating agents within the Collective Intelligence framework, and based upon recent developments in the Probability Collectives.

The novelty of this design methodology resides in:

  • the combined and simultaneous search for optimal control and design;
  • ·the coupling of Probability Collectives Theory with a model reduction algorithm issued from design with uncertainty. ·the use of instrumental concepts borrowed from the Complexity Theory associated with Swarm Intelligence;
  • the coupling of Probability Collectives Theory with a model reduction algorithm issued from design with uncertainty.

At a later stage and following a biomimetic approach, this design methodology and these control strategies will be implemented and tested on an innovative engineering system: a virtual school of fish serving as a computational design tool for a real-time distributed collision avoidance system for autonomous vehicles. The study of this virtual school of fish will allow us to investigate potentially non-natural cooperative emergent behaviors leading to robust and optimal collision avoidance strategies.

Fig. 1: Swarm topological neighborhood (r) versus Metric neighborhood (R)

Fig. 1: Swarm topological neighborhood (r) versus Metric neighborhood (R)

Fig. 2: Schematic of the underlying communication/interaction network in a swarm

Fig. 2: Schematic of the underlying communication/interaction network in a swarm

Fig. 3: Measure of global swarm behavior (agent alignment) for various information sampling times

Fig. 3: Measure of global swarm behavior (agent alignment) for various information sampling times

idc_HeritageBuildingsDigitalTectonic

Click image to view the poster