I am a researcher at the Robotics Institute at Carnegie Mellon University in Pittsburgh, Pennsylvania. My research interests include minimalist mechanisms, compliant manipulation, and legged locomotion. I am currently working with Professor Nancy Pollard on compliant dextrous manipulation.
I am interested in understanding the balance between computation and physical processes in producing behavior. Robots embody processes which connect computation, physical dynamics, passive stability, an environment, and humans. Well-chosen hardware dynamics can bring high performance and efficiency but usually at a cost of generality. Software is flexible but cannot overcome the limitations of physics. An insightful allocation between mechanism and computation can utilize the best of both to enable new robot tasks.
I approach this work by designing novel machines in pursuit of parsimonious design principles. My goal is to achieve the elegant simplicity which results from mechanisms well-integrated with the task dynamics. These machines are efficient and work with their environment rather than against it. They use a minimum of hardware and energy resources.
My work to date includes hopping robots using efficient compliant legs, walking robots inspired by passive dynamics, simple manipulators with externally steered compliant state, a low-cost snap-fit robot kit with an elastic neck, a simple force sensing palm, and kinetic sculptures using just a few actuators to create complex effects in fabric. All of these machines explore different tradeoffs between computing and physical dynamics to create behaviors.
I aim to develop further the principles of parsimonious design through experiments along several lines:
Programmable mechanism. Passive mechanisms configured by low-power actuators combine high bandwidth and energy capacity with the efficiency of orthogonal control inputs. As example, the Bow Leg uses low-power actuators to guide the high-energy natural dynamics of hopping. Similarly, series-elastic actuators can apply an elastic natural response to the task while still using geared motors.
Design methodology for rapid prototyping. New fabrication tools are quickly improving. These tools enable new design approaches for research robots, not just reducing time, but allowing forms which were previously unattainable. Taking full advantage of this requires rethinking our design vocabulary.
Direct tactile interaction with humans. Robots which communicate directly with touch are necessary for working closely as human partners. During shared tasks, the shared physical space suggests a natural interface vocabulary which is intrinsically task-related. The closest human analogy is partner dance: a leader and a follower closely coordinate movements using physical contact and trained responses which naturally guide each other. Future robots for human collaboration will need algorithms and physical form optimized to accommodate the dynamics of direct human interaction.
Art-friendly robotics. Artists are developers, users, and critics of technology. A deep understanding of systems thinking can connect research and the humanities for the benefit of both.
My art practice currently includes kinetic fabric sculpture which combines fabric structure, actuators, lighting, and sensors, and multimedia performances based on interactive transformative processes. I have also worked with talking mirrors which reveal contributed stories and secrets, plus a few miscellaneous projects and affiliations.
I am also a co-founder of Rossum's, a forum for artists working with robotics. This group was created by Ian Ingram and myself in order to bring together artists and engineers interested in robotic and mechatronic art. The format is an informal working group which meets regularly to discuss practical and conceptual issues in our works. The focus is on work which combines the digital and mechanical in embodied forms.
I am a graduate of the Carnegie Mellon Robotics Ph.D. program. As a graduate student I was a member of the Manipulation Lab, advised by Matt Mason and Ben Brown. For my doctoral thesis I developed planning algorithms to control a novel locomotion machine, the Bow Leg Hopper. The thesis document is available on my publications page.
"garthz" AT "cmu.edu" / PGP public key
Robotics Institute / Smith Hall 226 / Carnegie Mellon University / Pittsburgh, PA 15213 / USA
office: 412-268-5909 / home: 412-521-1282 / plan file
In old school web tradition, I have
some pages of useful links.
Some of my older content pages have moved to my archival page.
Last modified: 2013/01/19 11:13:53Garth Zeglin, Robotics Institute, Carnegie Mellon University.