Download presentation
Presentation is loading. Please wait.
1
T.J. Hartnett CSE 704 Seminar Dr. Miller Spring ‘11 EVOLVING CREATURES
2
AGENDA Creatures Structure Creatures “Nervous System” Genetic Algorithms ODE Problems (NaN and Normalization) The Linear Complexity Problem Future Work
3
CREATURES STRUCTURE Each “Limb” is a box with random size dimensions A Creatures Skeleton modeled as a graph Connections made with ball joints and hinge joints
4
CREATURES NERVOUS SYSTEM Neural Nodes Two inputs with random weights Add, Log, Integrate, Oscillate 20 Function total Each Limb had 6 Nodes associated with it Received input joint feedback and output as forces on limbs
5
GENETIC ALGORITHMS Random Initial Population of 100 creatures Fitness evaluation of motion The fastest survive and reproduce Reproduce by tweaking neural input weights, neural input/output connections, addition and removal of limbs
6
OPEN DYNAMICS ENGINE Models Rigid Body Dynamics in 3 Dimensions Includes Collision Detection World, Bodies, Mass, Geoms, dWorldStep Parallel?
7
PARALLELIZATION OF ODE Example Code….
8
LINEAR COMPLEXITY PROBLEM http://en.wikipedia.org/wiki/Linear_complementarity_problem Matrix M vector q LCP seeks vectors z and w which satisfy the following constraints w = M z + q W >= 0 z >= 0 (each component) W i z i = 0 M must be symmetric positive definite matrix (zTMz > 0 for all non zero real z vectors) ODE Problem with objects colliding, should be positive ODE ERP and CFM
9
REFERENCES "Evolving Virtual Creatures" K.Sims, Computer Graphics (Siggraph '94 Proceedings), July 1994, pp.15-22 "Evolving Virtual Creatures"
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.