Evolutionary computational synthesis of self-organizing systems

Abstract

A computational approach for the design of self-organizing systems is proposed that employs a genetic algorithm to efficiently explore the vast space of possible configurations of a given system description. To generate the description of the system, a two-field based model is proposed in which agents are assigned parameterized responses to two “fields,” a task field encompassing environmental features and task objects, and a social field arising from agent interactions. The aggregate effect of these two fields, sensed by agents individually, governs the behavior of each agent, while the system-level behavior emerges from the actions of and interactions among the agents. Task requirements together with performance preferences are used to compose system fitness functions for evolving functional and efficient self-organizing mechanisms. Case studies on the evolutionary synthesis of self-organizing systems are presented and discussed. These case studies focus on achieving system-level behavior with minimal explicit coordination among agents. Agents were able to collectively display flocking, exploration, and foraging through self-organization. The proposed two-field model was able to capture important features of self-organizing systems, and the genetic algorithm was able to generate self-organizing mechanisms by which agents could form task-based structures to fulfill functional requirements.

Evaluating the learning and performance characteristics of self-organizing systems with different task features

“…The box neighborhood is defined as six regions (Humann et al , 2014; Khani and Jin, 2015), as shown in Figure 11b. The box dynamics are based on the pymunk physics model.…”

“…Through agents’ local interactions, high-level system complexity can be achieved by a bottom-up approach (Reynolds, 1987; Ashby, 1991). Designing complex systems through a self-organizing approach has many advantages, such as adaptability, scalability, and robustness in comparison to traditional engineering systems with centralized controllers (Chiang and Jin, 2012; Humann et al , 2014; Khani and Jin, 2015; Khani et al , 2016; Ji and Jin, 2018). A swarm of robots is often homogeneous, with compact size and limited functionality, and is an example of such SOS (Kennedy, 2006).…”

Effect of Social Structuring in Self-Organizing Systems

“…Our previous work on CSO systems has provided useful insights into understanding self-organizing systems and introducing nature-inspired design concepts. A CSO system is based on some key concepts including a dDNA capturing all design information in a bit-string [1], parametric behavioral models for agent control and optimization [36,37], and an FBR mechanism [4].…”