We develop a real time simulation of varied swimming creatures in a fluid using Nvidia Flex for physically based simulation. Nvidia Flex uses position based dynamics and position based fluids to achieve real time performance. We model creatures' bodies as solid clusters of particles that are connected with joints and animated using muscles. A genetic algorithm is used to optimise muscle function in order to allow the creature as efficient forward swimming gait as possible. We present four different bodies and use our system to automatically generate their movement. We show that swimming in such a fluid is possible, albeit slow and it does not result in realistic turbulence. The evolved gaits are compared with those of real fish, namely anguilliform, carangiform and thunniform. The fit is different for different bodies.
|
