Simulating and Evolving Bipedal Locomotion.
I endeavored to write a computer program in C++ which implements a 2D biped (stick figure) with legs and a torso, in which the biped learns to walk and achieve other human motions. The biped simulates human actions by making tiny rotations of its limbs in small time intervals. The axiom that I assert in proposing the possibility of this project is that the biped can learn to do any human action by basing the subsequent tiny rotation of its limbs on the current position of its limbs. I keep track of position and motion in a large integer array, and improve the performance of the array with a genetic algorithm. Experimenting with different sizes and 'shapes' of the learning array, I have achieved my goal of the biped walking across the screen.
Jordan, Kalen, " Simulating and Evolving Bipedal Locomotion." (2000). URC Student Scholarship.
Ford Research Fellowship
This document is currently not available here.