engineering things
United Kingdom
Joined April 2015
- Tweets512
- Following507
- Followers791
- Likes17K
Pinned Tweet
Parking Simulator is now free to play! Updates coming very soon to bring more cars and levels to the game. Play now on PC, Phone, Tablet or Xbox!
roblox.com/games/5801575819/…
#robloxdev #rbxdev #roblox @ToppleGames
ok its fast again. this is a way more detailed voxel grid and way more explosions than would ever happen in game
average is 4ms and 99th percentile is 19ms which is like 50fps
only performance concern now is memory usage of all the meshes/textures on mobile i think
its a bit slow with this many explosions and chunk updates going on at once
i'll try profiling it and speeding up the slow bits next. it might also be able to run chunk updates in parallel but i've not tried that before
worst case i can just lower the voxel/render resolutions
I need to tune the noise, but adding material variations to the rock is pretty simple. I can use the marching squares to detect contours for each material and generates a few more meshes per chunk
Rendering the destructible terrain with meshparts. I'm honestly surprised the engine allows this.
Not being able to use a fragment shader makes texturing this so difficult. There's a few layers of dynamic meshes per chunk here to get the grass and mud layers to work.
voxels data -> chunks -> marching squares contour detection -> tracing edge loops -> convex decomposition -> ear clipping triangles
and its so fasttttttt and can update in real time!
mesh parts next. i really hope i dont hit an engine limitation so much work went into this haha
Grenades now go boom. I switched the terrain over to a custom marching squares approach. It's just creating thousands of parts right now. I will try out a dynamic mesh next + hopefully proper textures
Working on a 'parking sensor' feature for parking simulator that will detect when you're getting close to obstacles and show a warning + beep
Kinematic vehicle path planning is very difficult but I've finally made some progress.
This is using A* to explore the 'state space' and find a series of inputs that get the car to the target.
Its like pathfinding - but also considers momentum, turning circle, and collisions.
Making some progress on the vehicle collision predictor.
I've implemented an approximation of the physics engine, which can simulate the future position of car.
The next step is to try apply inputs based on this simulation to avoid the predicted collisions.
Very rough npc-vehicle controller. It doesn't pathfind or try to avoid collisions yet. It just tries to drive towards the target and park on top of it.
