Organisms, societies and computers are all complex systems whose behavior emerges from the interactions of components. For example, effective immune response requires coordinated interactions between billions of immune cells with no central point of control, and the behavior of ant colonies emerges from distributed communication between millions of ants. This research will examine how ant colonies and immune systems form distributed information exchange networks. By studying in detail how distributed interaction networks guide searching strategies in two distinct systems, this project will formulate a general theory describing how decentralized biological networks are organized to search, respond and adapt to different environments, and how they effectively scale up to large sizes. The broad theoretical goal is to understand how network size, structure and dynamics affect biological function.
The project will contribute to an interdisciplinary education program that integrates field experiments with computational modeling and serves pre-college, undergraduate and graduate students. Plans are presented to recruit and train members of underrepresented groups. The investigators are well situated to achieve this, as the University of New Mexico is one of only two universities in the nation that are both a Minority Serving Institution and a Carnegie Very High Research Activity University. Interactions with pre-college education projects will expose local students to interdisciplinary science in a way that is engaging and accessible.
Computation traditionally requires aggregating information in order to make calculations. However, complex biological systems have evolved mechanisms for decentralized information processing. This research demonstrates the use of decentralized communication, memory and movement in order to search in complex environments. We specifically demonstrate how ants use decentralized strategies when foraging for food and how cells of the immune system use decentralized strategies to protect against pathogens. We have built a robotic swarm--a team of cooperating robots--that interact with each other and their environment using algorithms based on decentralized biological behaviors. Our simulations tested how effective different foraging strategies were for resources that were distributed in different ways in the environment. Surprisingly, we find that colonies of ants foraging independently can sometimes be more effective than colonies of ants that communicate the locations of food using pheromone trails. When resources are dispersed at random, there is no benefit to ants that either remember or communicate information. However, when resources are clustered, the most effective foraging strategies balance individual memory and pheromone communication between ants, and adapt ant movement patterns based on cues they detect in the environment. There are many parallels between search by cells of the immune system and search by colonies of ants. Our simulations have focused on understanding how T cells search throughout the body and in specific organs to find and eliminate pathogens. Our models demonstrate that T cells combine fast movement through the circulatory system with slower but more thorough search in inflamed tissues to minimize the time to detect and eradicate infections. Our detailed analysis of T cell search in lymph nodes reveals that T cells combine directed, persistent motion with convoluted search paths in order to search both broadly and thoroughly. Our mathematical description of how T cells move demonstrates how they are able to detect pathogens in lymph nodes significantly more quickly than random Brownian motion. Like ants, T cells combine movement patterns with chemical communication and information sensed from the environment into effective search strategies. We developed algorithms based on these findings to control a team of iAnt robots, specifically built to collect resources using biologically-inspired behaviors. Evolutionary algorithms tailor those behaviors into foraging strategies that maximize performance under varied and complex conditions. The system evolves appropriate solutions to different environmental challenges, including increased communication when sensed information is reliable and targets are highly clustered, less communication and more individual memory when cluster sizes are variable, and greater dispersal with increasing swarm size. This work is the first to demonstrate that high-level behaviors can be automatically tuned to maximize the collective performance of robots to search varied and complex environments. This research sponsored by this grant has resulted in 22 publications, and several other manuscripts under review. The faculty and students involved in this research have given dozens of talks, including many for high school students and the general public.
