Research in the IRRL is focused on using robotics as a tool to answer interdisciplinary scientific questions. Current topics of active research include biorobotics, artificial learning, and evolutionary robotics. We have a number of robotic systems in the lab that have been designed by faculty and students and others that have been designed through work with industrial partners. Our robots include an underwater swimmer (Robot Madeleine), a group of Corobots, robots that evolve vertebral columns (Tadros), and a set of surface swimming robots (NERDs), among many others. The research in the IRRL has resulted in numerous publications.
Students and faculty working in the IRRL are involved in a wide range of projects, and we are always open to the development of new ideas and insights.
Can robots learn how to operate in the natural world with little to no prior information about what they will encounter? The projects in the lab that attempt to answer this question typically combine two goals: (1) discover high-level algorithms that contribute to our understanding of how newborn animals might accomplish this same task, and (2) develop ways to use those algorithms in the construction of autonomous systems that are useful to human beings. The most active project in the lab at the moment revolves around an effort to build simplified models of mammalian cortical columns that can be assembled in hierarchical architectures capable of discovering both concrete and abstract features of the world.
Biomechanics and movement control
Many of the projects in the lab take seriously the idea that the way a robot’s body is built has a lot to do with the kind of intelligence it can display. Projects in this category range from studies of how fish vertebrae might have evolved to solve certain problems of efficient movement, to the construction of RayBot, an underwater swimmer inspired by the Pacific electric ray.
Projects in this category often overlap projects in other categories because evolutionary algorithms provide solutions to a number of problems in robot design and development. The autonomous learning studies have used evolutionary algorithms to fine tune learning architectures, for example, and evolutionary processes were at the heart of the TadRo projects highlighted in the section on Biomechanics and movement control. One currently ongoing project in the lab began by using 10 swimming robots called NERDs (Neurally Evolving Robotic Devices) to explore how resource density affects control system evolution, but the project itself has evolved into a study of emergent properties of swarms.
One of the most difficult problems in robotics is how to share information between robots so that they can coordinate their actions in the performance of a shared task. Much of the work on this problem in the lab has been in simulation to this point, but the hardware exists to pursue these problems in robots.
Two projects in the lab have investigated various parameters of human control of robotic systems. The first investigated the marginal utility of increasing the amount of sensory information transferred from the robot to the operator. Does having more information allow for more effective, more efficient control? A second project involves the construction of a neural network to translate the movements of the operator’s arm into appropriate movements of the robot arm, even when the robot arm has a very different mechanical architecture from the human arm.
Science, Technology, Engineering, and Math (STEM) Education via robots
Working with robots requires many different skills and access to many bodies of knowledge, from physics and biology (and at times chemistry) to mathematics, computer science, and engineering. Faculty in the lab believe strongly in the value of robotics as an entre into these disciplines and have been active in the development of systems for enhancing STEM education using robotics.
An enormous amount of research in robotics can now be done in virtual environments thanks to the development of powerful new software tools for animation and for the simulation of the physics of the real world.