Talk about adaptation shape and behavior (AI and manufacturing)
- Self reflection and learning simulators
- Self assembly and programmable matter
Most of work initiated in robotics and expanded beyond it.
Industrial robots are super human in precision and almost all other factors compared to humans, but they lack adaptability.
iRobot roombas (vacuums) are more adaptable to a changing environment.
Cornell is looking at robotic evolution.
The Golem Project is discussed
Simulated system to learn and then created the robots and had them adapt to the environment.
It works for simple things but does not scale.
Hod Lipson website
Evolve in reality takes too many trials and is too expensive. Again it does not scale.
Co-evolution of simulator and robots
Simulate and get/evolve your best simulated robot
Then fabricate this robot and try it and get feedback sensor data
Then evolve your simulator to make a better simulated robot
and keep cycling through this loop.
Additional method- the learning system will determine a test for different choices and selects the test that would generate the most disagreement between choices.
- figure out what you have (how many legs, sensors etc...)
- then figure out how to achieve goals based on it
-adapt to losing a leg
Can it model its own controller ? Its own brain
Can a robot model another robot ? Social behavior
-theory of mind
-recursion (how does the other robot model the first robot)
Shape changing robots
(simple) Robots that replicate and assemble
Fabbing components for robots (print the components)
Goal is to fab a robot that will walk out of the printer (not yet)
Killer fab@home app is printing food.
Gizmodo today reports on a dedicated 3D food printer
MIT has the Cornucopia: Digital Gastronomy project
Cornucopia is a concept design for a personal food factory that brings the versatility of the digital world to the realm of cooking. In essence, it is a three dimensional printer for food, which works by storing, precisely mixing, depositing and cooking layers of ingredients.
Cornucopia's cooking process starts with an array of food canisters, which refrigerate and store a user's favorite ingredients. These are piped into a mixer and extruder head that can accurately deposit elaborate combinations of food. While the deposition takes place, the food is heated or cooled by Cornucopia's chamber or the heating and cooling tubes located on the printing head. This fabrication process not only allows for the creation of flavors and textures that would be completely unimaginable through other cooking techniques, but it also allows the user to have ultimate control over the origin, quality, nutritional value and taste of every meal.
Using millimeter sized bricks and components for assembling things.
Using reflection and sensing for a simulated bridge to figure out where a weak component was. It was faster and more accurate than traditional civil engineering methods.
Evolving equations to describe data.
Evolving equations to describe dynamics.
Evolve implicit equations for static equations that describe something about some data points.
Able to determine hamilitonians and lagrangians without knowing physics in advance.
Open source Eureqa has been applied to the stock market, patterns in herding cows and much more.
Hamiltonians are evolvable.
Accelerate our ability to hypothesize, test and model.
Others are working on evolutionary game theory