LightSwarm
Kyu Bok Lee, Palash Nandy, and Charles Rose, 2014

LightSwarm is an augmented reality swarm environment for testing human and swarm robot interaction. Overhead projectors create simulated agents in a physical space that can be inhabited by people and objects. Overhead sensors extract features of the environment such as people and elements for interaction. The end result of our project is a swarm of virtual fireflies in the space with people and showing some social behavior. (pdf) Watch the movie!

Pattern Formation by Evolution of Cellular Automata
Alex Garruss, Casey Grun, 2014

Biological systems are able to develop exquisitely complex patterns; the existance of very simple computer programs that can produce highly complex be-havior suggest that the rules underlying these natural pattern formation processes may be very simple. In this work, we attempt to use an evolutionary algorithm to explore the space of initial conditions for an elementary cellular automaton (Rule 110), known to be Turing complete. (pdf)

Evolutionary Design of Mechanical Linkages
Amanda Chassaei, Joy Ming, 2015

Mechanical linkages, or systems of rigid bars connected by hinges or other rotating components, have applications spanning the fields of robotics, animation, biology, art, and de- sign. However, designing linkages with desired motion is a highly unintuitive process. In this paper we describe the creation of a user interface for linkage optimization, our operationalization of the problem into an evolutionary computation algorithm, and two experiments assessing the creation of this workflow. (pdf)

Six Degrees of Collaboration: A Predator Pack's Guide to Success
Sami Ghoche, George Lok, Devvret Rishi, 2015

In this paper, we investigated different collaboration policies for a pack of predators chasing a moving prey, ranging from complete collaboration to no collaboration at all, using the Berkeley Pacman game platform. We operate in two settings: In the first setting, all predators can observe each other and their prey at all times, whereas in the second setting, predators can only observe what is directly in their line of sight. In the second setting, we present and compare two alter-native search and destroy strategies through stigmergic communication. In addition to evaluating the tradeoffs between mechanisms, we also present a successful trapping heuristic for a pack of coordinating predators (pdf)

Gradient-based Assisted Navigation and Control for Autonomous UAV Swarms
Bryan Kate, Peter Bailis, 2010

We present a system and corresponding algorithms for directing the flight of autonomous aerial vehicle swarms in order to perform long-distance navigation and control. Our solution relies on a static deployment of embedded devices in the environment; as the vehicle moves through the environment, it detects the presence of one or more artificial fields projected by the devices to compute a gradient. We characterize the feasibility of our system in both real-world tests (with berkeley motes) and simulation. (pdf)

Modeling Load-Adaptive Self-Assembly of Ants
Daniel Haas, Brett Harrison, Drew Robb, 2010

Ants can cooperate to accomplish complex tasks. If ants detect a food source that is unreachable due to spatial separation, ants can construct bridges out of their own bodies, across which other ants can travel in order to forage. While this phenomenon is well-documented by biologists, the mechanism by which this occurs is currently unknown. In this paper, we propose and simulate several models for ant bridge building, where ants extend the bridge one at a time, and send stigmergic signals to other ants when their local portions of the bridges become unstable. (pdf)

A Biologically-inspired Model of Organic Peer Assessment
Steve Komarov, 2014

We use a biologically-inspired model to analyze a type of peer assessment interaction between students. Bee nest site selection solves the problem of finding the best-of-n by a colony. Online facilitated peer assessment can also be framed as a best-of-n optimization problem. The goal is to use the model to shed light on a number of optimization questions, including: how many peer assessment participants are necessary to solve best-of-n, what should the student-to-solution ratio be, and how to counter the snowballing effect. (pdf)

Concept Learning for Regular Expressions
David Liu and Andrew Liu, 2014

Evolutionary algorithms harness ideas from biological evolution to explore complex search spaces. A recent variant called the E.C. algorithm aims to capture the search perspective of a human learner: the approach initially looks at simpler problems from the domain to learn useful and reusable solution concepts, which can then be applied to more difficult instances. The E.C. algorithm has previously been applied to problems such as symbolic regression and Boolean function learning. Here we use the E.C. algorithm to learn regular expressions for binary languages. (pdf)

Robust and Self-repairing Formation Control for Swarms of Mobile Agents
Jimming Cheng, Winston Cheng, 2004

We describe a decentralized algorithm for coordinating a swarm of identically-programmed mobile agents to spatially self-aggregate into arbitrary shapes using only local interactions. Our approach, called SHAPEBUGS, generates a consensus coordinate system by agents continually performing local trilaterations, and achieves shape formation by simultaneously allowing agents to disperse within the defined 2D shape using a gas expansion model. (Project led to a published paper (AAAI 2005)) Link

Firefly-Inspired Sensor Network Synchronicity with Realistic Radio Effects
Geoff Werner-Allen, Geetika Tewari, Ankit Patel, 2004

Synchronicity is a useful abstraction in many sensor network applications. Communication scheduling, coordinated duty cycling, and time synchronization can make use of a synchronicity primitive that achieves a tight alignment of individual nodes’ firing phases. In this paper we present the Reachback Firefly Algorithm (RFA), a decentralized syn- chronicity algorithm implemented on TinyOS-based motes. (Project led to a published paper (Sensys 2005)) Link