NICOP - Collective decisions in dynamic environments
Abstract
The CODE project looks at the processes leading to decision making in decentralised multi-agent systems(e.g., robot swarms), and specifically addresses complex problems that entail (i) uncertain and variable environmentalconditions, and (ii) spatial heterogeneities related to the (random) mobility of agents. Decisionmaking is a cognitive process that presents similar dynamics in systems as different as neural populationsand insect colonies, and is at the basis of more complex forms of information processing and behaviour(e.g., categorisation, task allocation). Besides the theoretical relevance, the principled understanding of thedecentralised mechanisms underlying decision making is fundamental for the design of large-scale artificialdistributed systems capable of flexible information processing, adaptive and robust behaviour in faceof changing conditions and external (possibly disruptive) events. In this way, it will be possible to providecognitive capabilities to large-scale distributed systems ranging from multi-robot to cyber-physical systemsincluding embedded devices, humans and robots. Such concrete application scenarios place important requirementsand constraints to the decision making abilities of large-scale artificial distributed systems, anddemand for a deeper theoretical understanding of the mechanisms and properties underlying the decisiondynamics.Stemming from the concept described above, this project will seek the following specific objectives.(i) Deepen the theoretical understanding of the mechanisms leading to optimal decision making in distributedsystems with respect to uncertain and dynamic environmental conditions. (ii) Identify and characterisethe interaction dynamics resulting from random mobility of agents under a wide range of conditions,including biases from the presence of other agents (e.g., attractions or repulsions), and study the influence ofthe mobility pattern on the diffusion of information within the system. (iii) Integrate decision-making andinteraction dynamics to identify the mutual relationship between the random mobility of agents and theoverlying decisional processes. (iv) Implement optimal collective decision making in large-scale artificialdistributed systems (e.g., robot swarms) to deal with dynamic environmental conditions and spatial heterogeneities.(v) Study the influence of external (possibly malicious) users trying to steer the collective decisionprocess, so as to devise human-swarm interaction methods that exploit the knowledge of the collectivedecision dynamics to maximise efficiency and account for resilience against targeted attacks.The CODE project will employ a mix of analytical, computational and experimental studies with robotswarms, to account for both abstract problems and concrete scenarios in which spatiality and heterogeneitiesare present. Analytical studies will be conducted to identify the possible system dynamics and relevantparameterisations. Multi-agent simulations will be developed to implement and test the microscopic interactionrules leading to desired system dynamics, as well as to deal with conditions too complex to be tackledwith an analytical approach. Links between microscopic and macroscopic descriptions will be drawn, soas to seamlessly move between the description of the individual rules governing the agents behaviour andthe system-level dynamics. Robotics experiments provide a mean to test the ability to design artificialdistributed systems based on the gathered theoretical knowledge.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 10, 2018
- Source ID
- N629091812093
Entities
People
- Vito Trianni
Organizations
- Consiglio Nazionale delle Ricerche
- Office of Naval Research
- United States Navy