Multi-objective control and coordination of multi-agent systems

Multiattribute utility copula;multi-objective control;multi-agent system;multi-robot testbed;dynamic coverage control

Valicka, Christopher

关键词: Multiattribute utility copula;    multi-objective control;    multi-agent system;    multi-robot testbed;    dynamic coverage control;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/46676/Christopher_Valicka.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

In this dissertation, control methodologies for systems of multiple mobileagents facing multiple objectives are considered. Carefully constructed objectivefunctions for agents interested in any or all of collision avoidance,robust communications, waypoint following, and dynamic coverage objectivesare presented. Theory relevant to gradient-like control laws designedusing over-approximations of the maximum function is studied and applicationsof developed controllers in simulations and physical experiments arepresented.Agent behavior resulting from and relevant to the various objectives isdiscussed. Amongst the various objective functions, novel formulations forproximity and global coverage objectives are presented. A novel control lawdesign is developed based on copula structures from multiattribute decisiontheory. Systematic and intuitive methods for determining the tradeoffs betweenmultiple objectives are presented and applied to the design of multi-agentcontrol laws.Using over-approximations of the maximum function, appropriate definitionsare presented to develop the conditions required for an agent to accomplishmultiple objectives. The proposed control laws' suitability for andability to be implemented straightforwardly on various multi-agent systemsfacing multiple objectives are demonstrated in multiple simulation and experimentalresults. The proposed copula based control law and global coverageapproach are discussed and validated through their combined implementationin a multi-robot testbed.