Jams

randompast

My recent research has lead me into 2nd order differential equations (heat, laplace, schrodinger, fokker-planck, etc) which have applications in a wide variety of domains.  The initial attempts I had were in line with the explicit Euler Method.  Currently, I'm learning about stability analysis and implicit representations.  I've always seen games as interactive simulations and really have enjoyed emphasizing the simulation side of things, this takes me a step further in that direction.  Here's a very cool example for Laplace's Equation in Python <https://www.youtube.com/watch?v=dKCAVteveYc>.  Now, if you will, imagine this creating field forces for a physically inspired game (gravity, waves, stresses).

I wanted to make a neuron based simulation game.  Previously, I had tried a pendulum with sliders representing synaptic weights.  Since the solution was an unstable equilibrium (balancing upside down) I figured it'd be a lot more difficult than simple movement.  So, I went back to the drawing board to see if I could create something like a virtual fish.  The graph dictates the behavior with a simple 2 neuron setup where the angle difference towards the goal (green dot at 2,2) is fed as input to the neurons, each with a different kernel.  I've turned off the neuron turning and am instead turning via a normal distribution σ=0.1.  The question that's causing me difficulty is "What should the neuron's function be to cause it to aim towards the goal?"  That is to say, I need a way of transforming the angle history through a neuron (filter and ahp) to cause impulse torques to turn towards the goal.  Graph description: path (blue, starting at 0,0), neuron 1 (red), neuron 2 (lime), goal (green 2,2).

Added a reset button for the pendulum, and interneuron communication.  I still can't "beat it", not that it has an official win state, but the goal is to get it to balance upside down.

Managed to get a shader graphing the angle, angular velocity, and angular acceleration using uniform variables.

It's a proof of concept for a pendulum balancing neuron game.  The design is really dumb right now, advice is welcome.  I was going to try to do this for the create-a-thon at badass, but I decided I wanted to sleep instead, then thought maybe Ludum Dare?  It'd sure be nice to use numpy in Godot.  I've been feeling stifled not having it.

Jams

Dev Logs