A design for dynamic self-organization in autonomous distributed operations

Abstract

Considerable research has focused on the development of single, autonomous vehicles - vehicles that are able to sustain themselves and achieve specific objectives, but that generally lack the ability to autonomously coordinate with other vehicles. As this research evolves into operational systems, a key challenge is to develop autonomous vehicles that can self-organize in order to collectively address a set of goals that could not be achieved by any one individual. This thesis develops a framework for an operational system called Autonomous Distributed Operations (ADO) and presents a design for self-organization within this framework. Self-organization is shown as an integration of five activities that include goal evaluation and adoption, organizational strategizing, goal decomposition and aggregation, organizational design, and finally, organization adoption. A detailed implementation of the fourth component - organizational design - is presented. A four-phase systems engineering methodology is developed for designing organizations by sequentially optimizing along multiple dimensions, which include boundary, size, structure, and membership. Algorithms that yield optimal solutions for each phase of the design methodology exist, but are not suitable for realistic scenarios, which are often characterized by large state spaces. Therefore, the design methodology is implemented using heuristic algorithms that trade optimality for problem tractability. Results of the design process are presented using a variety of goals and scenarios as inputs. The scenarios are intended to reflect situations commonly encountered in Military Operations in Urban Terrain.