Date post: | 07-Apr-2017 |
Category: |
Technology |
Upload: | reem-alattas |
View: | 68 times |
Download: | 1 times |
Evolutionary Robotics
Introduction
• Evolutionary Robotics is a method for the automatic creation of autonomous robots.
• Inspired by the Darwinian principle of selective reproduction of the fittest captured by evolutionary algorithms.
Ref: Handbook of Robotics - Chapter 61: Evolutionary RoboticsDario Floreano, Phil Husbands, and Stefano Nolfi
Evolutionary Roboticsand Open-Ended Design Automation
Hod LipsonCornell University
Systematic Synthesis
• A rational approach to synthesis is needed.• The process of successive adaptation by
improvement and recombination of basic building blocks.
• It is open ended, unlike classical genetic algorithms.
Optimization vs Synthesis
Optimization• We tune the values of a set of
parameters in order to maximize a target function. The set of parameters, their meaning and their ranges are predetermined.
• E.g. To design a circuit, we manually provide a basic layout of resistors, capacitors, and coils, and then try to automatically tweak their values so as to maximize performance.
Synthesis• An open-ended process where
we can add more and more components, possibly each with their own set of parameters.
• E.g. To design a circuit, we could start with a bucket of components, and use an algorithm to automatically compose them into a circuit that performs the target function.
A Simple Model of Evolutionary Adaptation
• Initial Population: a large set of initial candidate designs.• Perform repeated selection and variation:
– To perform selection, we first measure the performance (fitness) of each solution in the population.• The fitness metric needs to be solution-neutral, i.e measure the extent to
which the target task has been achieved, regardless of how it was achieved.– We select better solutions (parents) and use them to create a new
generation of solutions (offspring).• Offspring: are variations of the parents, created through
variation operators like mutation and recombination. • The process is repeated generation after generation until good
solutions are found.
Robot Body & Brain
• Body Morphology Hardware• Brain Controller Software– Evolving Controller (fixed morphology)
• Legged robot simulation and physical experiment– Evolving Bodies and Brains
• 3D cubes and oscillators experiment
• With design automation, we could reverse engineer the evolved controller to find out exactly how it works
Morphology Representations
• Genotype: encodes information for growing, or developing, a phenotype.
• Tree Representation• Developmental Representation• Regulatory Network Representations
Tree Representation
• A set of operations to construct a phenotype in a top-down or bottom-up manner.– Top-down representation starts with an initial
structure (an embryo) and specifies a sequence of operations that progressively modify it into its final form.
– Bottom-up construction of a one-DoF mechanism begins at the leaves of the tree with atomic building blocks and hierarchically assembles them into components.
How could a tree-representation be used to describe robot morphologies?
• Top-down construction of a mechanism starts with an embryonic kinematic basis with the desired number of degrees of freedom (DoF’s).
• A tree of operators then recursively modifies that mechanism by replacing single links (DoF = -1) with assemblies of links with an equivalent DoF, so that the total number of DoF remains unchanged.
Developmental Representation
• Allows the robot’s morphology to develop from a basic “seed” and a set of context-free development rules.
• L-system is a set of rules like the “AB” and “BAB”. These “rewrite” rules are parametric (i.e. may pass parameters), and have conditions (are executed only when the parameters meet some conditions).
• An evolutionary algorithm was used to evolve individual L-systems, that when executed produced a build sequence which produced the machine.
Regulatory Network Representations
• Robot elements; such as actuators and sensors, are connected through a neural network, but the specific connectivity of the network is determined by an evolved regulatory network.
Evolving Machines in Physical Reality
• Evolving controllers for physical morphologies:– To make a perfect simulator, that whatever works
in simulation will work in reality equally well.– To use a crude simulator that captures the salient
features of the search space.– Co-evolve simulators so that they are increasingly
predictive.• Exploration phase: evolving the controller.• Estimation phase: creating the simulator.
Evolving Machines in Physical Reality
• Making morphological changes in hardware:– Reconfigurable Robots: are composed of many
modules that can be connected, disconnected and rearranged in various topologies to create machines with variable body plans.• Self-reconfigurable robots are able to rearrange their own
morphology, and thus adapt in physical reality.• Producing the entire robot morphology automatically using
3D printers. This “printer”, when coupled to an evolutionary design process, can produce complex geometries that are difficult to produce any other way, and thus allow the evolutionary search much greater design flexibility.
Similar Topics
ModularSelf-Reconfigurable Robots
• A robot that changes its geometry based on task.
• Daniela Rus, MIT.– Origami Robot – MIT: http://
newsoffice.mit.edu/2014/mobile-folding-robots-0807
– Robo-Cubes – MIT: http://techcrunch.com/2013/10/04/m-blocks/
Self-Replicating Repairing Robots
• A machine that can build copies of itself.• Hod Lipson, Cornell.– http://
www.news.cornell.edu/stories/2005/05/researchers-build-robot-can-reproduce
Now, he is working on Evolutionary Robotics!
But, this one…
Robots Use 3D Printing and Simulations to Evolve
• Oslo University: http://www.ign.com/articles/2014/11/13/robots-use-3d-printing-and-simulations-to-evolve
We need to start with…
Evolutionary Algorithms (EA)
• EA are stochastic search and optimization heuristics derived from the classic evolution theory, which are implemented on computers in the majority of cases.
Classification of EA Methods
Tools
GEATbxGenetic and Evolutionary Algorithm Toolbox for Matlab
Darwin2k
• Simulation and Automated Synthesis for Robotics– http://darwin2k.sourceforge.net/
Thank You!
Q & A