 |
|
|
|
|
|
Malone, Lipson and Pariard
|
|
|
|
Malone, Lipson and fab@home |
Lipson and his fab@home team seek to address the grand challenges of Robotic Evolution.
Hod Lipson didn’t set out to revolutionize manufacturing. He just wanted to design a really cool robot, one that could “evolve” by reprogramming itself and would also produce its own hardware—a software brain, if you will, with the ability to create a body. To do this, Lipson needed a rapid-prototyping fabrication, or “fabber.” Picture a 3D inkjet printer that deposits droplets of plastic, layer by layer, gradually building up an object of any shape. Fabbers have been around for two decades, but they’ve always been the pricey playthings of high-tech labs—and could only use a single material.
Some day, Lipson believes, every home will have a "fabber," a machine that replicates objects from plans supplied by a computer. Such devices could change how we acquire common products, he suggests: Instead of buying an iPod, you would download the plans over the Internet and the fabber would make one for you.
“To really let this robotic evolutionary process reach its full potential,” says Lipson, a Cornell University computer and engineering faculty member, “we need a machine that can fabricate anything, not just complex geometry, but also wires and motors and sensors and actuators.” Lipson and his grad student collaborators, Dan Periard and Evan Malone, decided to put the problem to the people. They developed a low-cost, open-source fabbing system—Fab at Home—and encouraged experimentation by starting an online wiki for hobbyists. People report printing with everything from food (Easy Cheese, chocolate), to epoxy, to metal-powder-impregnated silicone to make conductive wires.
A Fab at Home kit costs around $2400. Lipson compares it to early kit computers such as the MITS Altair 8800, which democratized computer technology in the 1970s. At-home fabrication, Lipson says, “is a revolution waiting to happen.” As for that robot? Wait a year, he says, and it really will walk out of the machine.
- Adapted from articles by Logan Ward in Popular Mechanics and Bill Steele in the Cornell Chronicle
|
“To really let this robotic evolutionary process reach its full potential we need a machine that can fabricate anything, not just complex geometry, but also wires and motors and sensors and actuators.” Prof. Hod Lipson |
About Lipson's Research:
The fab@home project is an outreach component of Lipson’s research in the area of robotic shape adaptation, which attempts to address one of two “grand challenges” of engineering: (a) Can we design machines that can design other machines, and (b) Can we make machines that can make other machines. Both of these questions lie at the crux of understanding the engineering process itself, and progress on these fronts can offer huge leverage in our ability to design, make and maintain increasingly complex systems in the future.
|
|
Golem, the first evolved, printed and functional robot.
| |
Robotics: Shape adaptation through Multimaterial 3D printing and reconfiguration
Animals achieve their physical robustness not because they are made of highly durable materials, but because they are able to physically adapt and self-repair. In contrast, most efforts in robotics have focused on adaptation of controllers in a mostly fixed (unchanging) morphology. In the Golem Project (left), Lipson and Pollack showed the first evolved robot (both morphology and control) that was printed and physically functional. The question of how can machines adapt their shape is fundamental both for improving long-term machine robustness as well as for exploring and exploiting brain-body co-adaptation. These projects explored two technologies that can be used to realize these new directions. The first is a multi-material 3D-printing process for freeform fabrication of active systems such as actuators, batteries, conductors and structure and the second is a scalable, programmable modular self-assembly at macro and micro scales based on stochasticity. These two technologies also have applications well beyond robotics. |
|
Printable machines
(With Evan Malone). This project developed a multi-material 3D-printing process for freeform fabrication of active electromechanical systems. including sensors, actuators, power sources and logic, all co-fabricated simultaneously – not unlike biology. The main challenges in achieving this goal are development of functional links – mutually-compatible materials with appropriate rheological properties and desired functionality. Combined with suitable multi-material deposition processes like inkjet and feed deposition, we have demonstrated the first freeform fabrication of batteries (zinc-air and Li-ion), actuators(conductive polymers and ionomeric polymer-metal composites) that are low-voltage, air-operable and mutually compatible, strain gauges, wires, and flexible joints, all fabricated on the same system. A impending milestone is to fabricate a complete integrated system, such as a robot that could walk out of the printer.
|
|
|
|
3D printed active components: Structure and flexible
joints, embedded strain gauges, a battery and IPMC
actuator, all printed on the same system | |
|
|
The physical instance of machine self-replication
| |
(With Victor Zykov, Stathis Mytillianos and Bryant Adams). Morphological adaptation can also be achieved by exploiting modular substructure. We have recently demonstrated a new modular robotic system capable of complete self-reproduction for two full generations, as self-reproduction is an important form of autonomous self-repair and key to long-term evolutionary adaptation in biology. We have also put forward new mathematical framework that treats self-replication from an information theoretic perspective, yielding a quantitative metric of self-replication that challenged the classical von Neumann model that treats self-replication as a ‘binary’ property that a system either has or has not, and thus does not capture the rich variety of replication forms. |
|
Stochastic modular reconfiguration
(With Victor Zykov, Michael Tolley, and David Erickson). One of the main challenges in the field of modular robotics is scalability to large numbers (thousands) of units, an ability that will provide true flexibility in shape and form. There are fundamental power and precision limitations on the number of units that can be manipulated into place using conventional architectures. Many of the scaling issues could be overcome if the modular units could be reduced to microscale, but then many of the actuation and control issues are exacerbated due to ambient Brownian noise. The alternative concept of stochastic self-reconfiguration that we developed exploits, rather than tries to avoid, the natural stochastic noise of micro-scale dynamics. According to our approach, dormant solid state units float neutrally in an agitated fluidic environment, and are attracted or rejected into place by the main body according to some program, by varying local fluid flow patterns. This concept will allow structures to selectively grow, reject and reconfigure material in a programmable way by exploiting transport phenomena in their environment, not unlike biology. We have studied this concept in simulation and then physically demonstrated it first at macro-scale in 2D, then at macro-scale in 3D in an oil tank.
|
|
|
|
Dynamically reprogrammable fluidic
self assembly & reconfiguration | |
|
