The Evolution Park provides a testbed for applying evolutionary computation to the development and control of autonomous robotic systems. Mixed populations are created, including some robots and some natural systems to facilitate a form of biomutualism where biology, engineering and computer science inform one another in synergistic and mutually beneficial ways. The 3D printer enables the realization of physical bodies (morphologies) that evolve concurrently with their control systems. The fabricated bodies are coupled with electroactive polymer materials to produce artificial organisms capable of locomotion without motors. Specially instrumented aquatic environments allow robotic fish to be used as stimuli to elicit behavioral responses in living fish under conditions manipulated by the experimenter. A collection of high-performance graphics workstations with large monitors facilitate fine-grained, interactive analysis of evolved behaviors in simulated robots, as well as analysis of video data captured by underwater cameras.

This infrastructure enables compelling new research directions in evolutionary robotics and squarely builds on previous work by the PIs on group communication and cooperative behavior, parallel processing, wireless networking, autonomic computing, distributed algorithms, high-assurance software, complex behavior in natural organisms, sensors and actuators, robotic fish, bio-inspired robots and swarms, as well as speciation and the evolution of communication. These advances benefit many applications in science and engineering, public safety, and national defense. The Evolution Park supports several innovative educational and outreach activities. Such endeavors provide a framework for hands-on experiments that are integrated into university courses, teacher training workshops, summer camps and after-school programs for K-12 students.

Project Report

Motivations. Increasingly, mobile robots are aiding humans in remote and sometimes dangerous applications. Examples include monitoring natural environments, inspecting physical structures for faults, searching for disaster victims, and assisting in hazardous occupations such as mining and firefighting. Despite many advances in the past decade, however, most robots do not approach the performance of biological organisms: the materials, sensors, and actuators that comprise these systems simply are not as effective as biological tissue. The primary goal of this project was to construct Evolution Park, an experimental testbed for the design, fabrication and testing of autonomous robots that can address and overcome challenges of operating in the physical world. A major theme of the multidisciplinary facility is to enable engineering, biology, and computer science to inform one another in synergistic and mutually beneficial ways. A particular research focus of the principal investigators is robotic fish, which accomplish swimming by deforming their bodies or fin-like appendages. Such robots are intended to track oil spills, to measure water quality, and monitor underwater structures. In addition, biologists can use these robots to elicit behavioral responses in live fish in order to better understand these creatures and how they are affected by changing ecosystem. Research Facilities. The Evolution Park project expanded a robotics testbed established with an earlier NSF grant. The testbed now includes the following main components. First, a multi-material 3D printer enables the fabrication of robot components, which are then coupled with microcontrollers, motors and batteries to produce fully functional robots capable of locomotion and complex maneuvers. The printer is capable of jetting multiple materials simultaneously, enabling printing of assemblies that combine rigid and soft elements, such as fins and wings with rigid struts and pliable inner membranes. Second, specially instrumented aquatic environments, including a a 4500-gallon tank, swim tunnel, and motion capture system, enable evaluation of robotic fish prototypes, as well as experiments combining robotic fish and live fish. Third, a collection of commercial terrestrial microrobots is used to test adaptive control algorithms. Finally, a collection of high-performance computers supports the design and evaluation of robots in simulation and the analysis of video data captured by underwater cameras. Evolutionary Robotics. One of the main design approaches applied in the laboratory is evolutionary robotics. Evolutionary algorithms codify the basic principles of genetic evolution in computer software. In evolutionary robotics, an artificial genome encodes a robot's control system and possibly aspects of its physical structure. A population of simulated robots evolves over generations, with the fitness of an individual calculated with respect to performance on one or more tasks. Over generations, the bodies and controllers of the simulated robots evolve to improve performance on the task. The most successful systems can then be fabricated and tested. The Evolution Park testbed has enabled researchers to extend evolutionary robotics into new areas, including evolution of controllers that exploit passive and flexible materials, evolution of robots that effectively balance performance and energy efficiency, and self-modeling algorithms that enable robots to dynamically evolve compensatory behaviors when damaged or facing unexpected situations. Supporting Broader Communities. In addition to supporting the research of the principal investigators, the Evolution Park facilities have enhanced research and educational activities across Michigan State University. In the College of Engineering alone, the 3D printer has supported the research of 26 different faculty members as well as numerous undergraduate students, graduate students and research staff. Applications vary widely and include fabrication of components for hopping and flying robots, tiny structures for manipulating nanodevices, microfluidic components for cell culturing, neuroscience apparatus for monitoring animal behavior, components of high-frequency antennas and electronic circuits, and even three-dimensional fingerprints. The Evolution Park facilities have been used in six undergraduate courses and several senior design projects in Electrical and Computer Engineering, Computer Science and Engineering, and Mechanical Engineering. The testbed has also been integrated into MSU's Research Experiences for Teachers program, a popular summer program offering research opportunities to teachers from middle schools and high schools. These teachers translate their experiences into lesson plans for use in their classrooms as well as distribution to other schools. Finally, the 3D printer, water tanks, robotic fish, and terrestrial robots have greatly aided efforts to engage K-12 students and the general public through numerous outreach programs.

Agency
National Science Foundation (NSF)
Institute
Division of Computer and Network Systems (CNS)
Type
Standard Grant (Standard)
Application #
1059373
Program Officer
jeffrey trinkle
Project Start
Project End
Budget Start
2011-04-01
Budget End
2014-03-31
Support Year
Fiscal Year
2010
Total Cost
$305,000
Indirect Cost
Name
Michigan State University
Department
Type
DUNS #
City
East Lansing
State
MI
Country
United States
Zip Code
48824