Been a while since I last updated. I assure you I have not been resting. Unfortunately I am working on projects that need to remain under the cover of vagueness until they are either 1)more fleshed out or 2)officially launched. But that doesn’t mean I can’t tease a bit, right?
The two things that have been occupying most of my time the last couple weeks are procedural landscape generation and population, and procedural plant generation. The landscape project started because of an installation the Barbarian Group is doing for a gallery in Seattle (nope, not the McLeod Residence this time). This project should be done sometime in October. I am going to be vague on the details because there is still plenty of time for the scope to change and knowing me and the way I like to work, it will change. Four months is a long time.
There are some additional test renders up on my Flickr page. In short, you are looking at a terrain mesh based on actual GSPS data from NASA. Many thanks to Kyle McDonald for figuring out how to parse the very strange .hgt file format. His example code is available at OpenProcessing.org.
I have done terrain experiments before but never thought much about how to make them more lush by populating the mesh with plants and trees. Thanks to my new found appreciation for GLSL shaders, I was able to put a nice coating of wind-blown grass onto the terrain, as well as a few thousand trees and bushes.
The water is made of layers instead of a flat plane. This was mostly an aesthetic decision. I just like the way it looks, especially if the camera is lower to the ground as it will be in the final version.
It is interesting to pause here and consider how far we have come since the good ol days of the NextFest grass wall project. That project from 2006 stretched my coding ability as far as it would go. And all for what? A few hundred 2D grass blades that barely broke 30 frames per second. This new landscape has rolling hills of seemingly millions of blades of grass all bending in the wind with cloud shadows and minor dynamic lighting and easily hovers above the 60 fps threshold. Exciting! But thats pretty much all I can say about that until we get further along.
I recently got sidetracked because I wanted to find a substitute for the TurboSquid.com textures I was using for the trees and bushes. I remembered the branching application I wrote earlier this year and decided it would be reasonable easy to recode it to produce plant life.
I was partially right. It wasn’t hard to make the code churn out plants, but it was hard to make it create plants that didn’t look like CG plants from The Lawnmower Man. There was just way too much symmetry and predictability in the growth patterns. I decided to spend a few days making it more robust.
First up, trees. The image below was my attempt to make a believable tree (without branches). The basic process is to start with a node. Think of it as the seed for the tree. The seed is created with all the behavior characteristics for the entire tree. The seed creates child nodes which pass along the tree’s ‘genetic information’. This information consists of parameters like lengthDelta, lengthDeltaDelta, radiusDelta, radiusDeltaDelta, maxChildren, color, maxGenerations, etc. As a parent creates a child node, it sends this information but mutates it slightly. The nodes arrange themselves in space using magnetic repulsion. Any node can repulse other nodes as long as they are of an equal or higher generation. The node at the base of the trunk repels everything whereas the nodes at the branch tips repel only each other. If you code in some decreasing branch lengths and radii as you go from generation to generation, you will create a nice space-filling tree with no branch overlaps. Sadly, these trees are complex enough to elude a reasonable frame-rate but perhaps with some limitations placed on the number of branches that can be created, and killing off the repulsion after the branches settle into place, they will be swaying in the Perlin noise breeze in no time.
I then tried to place some leaves or flowers on the trees and accidentally changed the perceived scale quite a bit. It no longer seemed like a majestic oak. It turned into a bush, and then eventually, much smaller flowering weeds. I wasn’t put out by this change (bushes and weeds were going to need to be addressed eventually anyhow) so I went along with it and started making bushes and weeds.
The weeds ended up catching my attention the most because I could see a great deal of potential to create amazing unique renders of realistic looking plants without needing to manufacture them on a more traditional 3D application.
I was reminded on Flickr that I am treading into territory that has been explored by countless others before me, namely Prusinkiewicz and Jon McCormack. It is definitely an exciting distraction and I look forward to experimenting further with this methodology.
