Scribd
Upload
28 views2 pages

Evolutionary Algorithm

This document describes an evolutionary algorithm to optimize robot gaits. The algorithm begins by initializing a population of robot gaits and evaluating their fitness by testing on a real robot. It then selects the next generation by choosing the best performing gaits and mutating their poses. This process repeats, evaluating and selecting new generations, until a termination criteria like reaching maximum generations or a fitness threshold is met.

Uploaded by

Matthew Leccadito
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
28 views2 pages

Evolutionary Algorithm

This document describes an evolutionary algorithm to optimize robot gaits. The algorithm begins by initializing a population of robot gaits and evaluating their fitness by testing on a real robot. It then selects the next generation by choosing the best performing gaits and mutating their poses. This process repeats, evaluating and selecting new generations, until a termination criteria like reaching maximum generations or a fitness threshold is met.

Uploaded by

Matthew Leccadito
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Evolutionary Algorithm

start

initialization

1
termination
done
critera met?

select next
generation

mutation

evaluate fitness
initialization
 instantiate array of Gait - members of the population, suggested name: Robots
o load Robots with initial pose for all steps
o store Robots in file (we'll talk about this later)
 generation counter = 0
 fitnesses to 0

evaluate fitness
 send poses to real robot for fixed time period
 note distance from starting point and provide to program
 repeat for all Robots
 compute average fitness
 store Robots in file

select next generation


 select Robot with best fitness (elitism)
 choose several—about 30% of the total—Robots at random and compete, i.e., compare
fitnesses (tournament selection)
 select Robot with best fitness in the tournament
 repeat the tournaments until all members of the next generation are filled
 bump generation counter

mutate
 for each Robot in the new generation, for each pose, for each leg
 draw two random numbers from range 0 to 1
 add the two numbers and subtract 1
 multiply by the current maximum mutation angle
 add to the azimuths with limits to the max/min angles
 repeat starting at drawing two random numbers for the elevations

termination criteria
 maximum number of generations, OR
 some predefined level of fitness

You might also like