Dia 11: Bienvenidos a Arroyo Seco

Done (For Now)

After finishing my level, I spent a couple days catching up with 'day job' issues and updating my venerable copy of Sony Vegas to edit the footage. Despite missing my elapsed time goal, I built a complete finished level in about two weeks of spare time during the busiest time of the year . Not only do I now have something to show my peers for the many hours I've spent working in T3D, but the level has a dual use as 'eye candy' to help stimulate crowdsourcing and other fundraising strategies.

I am now ahead of myself in the sense that I have to add script and code functionality before I can do much more to the level. This will help me prioritize the tasks I need to finish my project on time. Against the backdrop of the finished level, issues like missing animations, missing debris files, sounds, gui, etc. are more obvious and thus will be worked on first. This 11-day challenge helped to break me out of a slump I'd fallen into with my AI development. With a sense of accomplishment behind me, I have new ideas I'm already implementing.

Dia 11: Finishing Up

Late night/early morning playtesting through the level revealed a few places where the engine would 'stutter'. Day 11 began on a roll, as I found and replaced the culprit objects. The bowls I was using as censors were a whopping 4k polys. A quick blast of ProOptimizer in MAX brought them down to a couple hundred, LOD'd down to a handful. I'd also left some stray, almost-imperceptible faces in the pyramids I'd modded. Holes fixed, faces cleaned, I got a few more fps out of my toil. I clocked a good 50+ fps with about two dozen AI running, with VS2012, MAX, Word, Torsion, Gimp and Fraps all open and running. Toggled over to debug to double-check my performance and BOOM! Glorious, magnificent lock-up crash...

Deconstruction

Hours later, I'd taken the whole level apart and found that some of the legacy TGE 'placeholder' models I was using were squeaking by in release but debug was invalidating them. This time, quickly re-imported into U3D, cleaned up in MilkShape and onward. In the process of tracking them down, Sean helped me track a second issue, a bug that had been hiding in some deprecated code for a while. Despite the frustration of time lost, I gained quite a bit in performance and reliability in the process. The skybox was slowing it down for some inexplicable reason, and with another, more ominous looking one, the level was finally running smoothly with no errors.

Pathfinding

I have been using multiple navmesh objects in my AI work for a minute, and Walkabout made setting them up much, much, easier. After RTFMing a couple more times, I setup a practice level and played with different sized terrain geometry vs. navMesh settings untill I found settings that would work. Once the meshes I needed were built, it was pretty painless to adapt my existing code/scripts to work with the new commands. Even my fledgling mech class with broken animations was target-seeking and acting like 'real mechs' in no time. I couldn't help but add Steve Acaster's 'Bombardment' resource, and much to my delight, my vision of a war-torn landscape had come alive. I'd found myself facing 8 mechs, 12 dragons, 8 flyers, and a phalanx of 12 alien troopers with my fps still running high 40s. Still a long way to go engine/resource/texture-wise, but my level was holding up and looking good.

Successes

I have learned that just as the painter readies his brushes and oils, I need to optimize my assets and pipeline before beginning the level design process.

• planned, built and delivered a playable level in spare time during the busiest two weeks of the year.


• used the map to integrate existing AI code with Walkabout


• setup map to be used with Sahara when features are implemented


• learned more about using and optimizing materials, models and other assets in T3D


• coherent color scheme helps player know his position by subtle suggestion of geographic/organic positions as well as to contrast bright colors of gameplay elements - actors, vehicles, weapons, particles.

Challenges

• Re-rigging models, combining/compressing textures and other asset preparation cost days of level development time


• Terrain 'stitching' needs to be addressed and code changes made, if needed.


• 'Feature Creep': Adding Sahara, Walkabout and other gameplay functionality increased development time. Knowing when to stop adding and start finishing is difficult.


• 'close up / fps' and 'far out / vehicle' views may need to be re-evaluated for performance.

Conclusion

Success breeds success. Having finished a level, I now have momentum and fresh energy with which to meet my next challenges. Its also forced me to address some issues I'd become desensitized to. It's easy for any digital creator to get so 'tunnel visioned' into the small details they're working on and overlook glaring problems.

And now, here's the vid! I obviously have quite a bit of animation/particles/damagedebris/sound to work on, but the immersion and gameplay I envisioned are already there.


Day 10