Gridless: New Horizons - An open world driving simulator with *somewhat* realistic physics and car models. The game drops you into an endless procedurally generated city with a full day/night cycle and live traffic lights. Drive freely, weave through traffic, and earn cash for close calls and reckless speed - then spend it in the Garage to unlock new cars.
Devlog 7 - Today I focused on learning how to make a minimap, this was quite a long process as I watched a few YouTube videos on how to use the Renderer function, however it did come out quite well in the end, as seen in the picture it shows a sort of simplified top-down view of the road and also smaller versions of the AI traffic on the roads.
I also tried to build functional traffic lights at each intersection. I had to use the inbuilt tools to model the lights as I couldn’t really find any proper pre-built models. They do seem to work in terms of changing colours, but I am yet to pair it with the traffic behaviour which I will be doing soon. I also ported in a few more Kenney Kit car models to upgrade the selection of driveable cars in the game. I will soon start the currency system and law system as well.
Dev log 6-
WE HAVE PROPER BUILDINGS NOW!!
HEAVILY improved the physics model. The physics model is arcade-style, built around per-vehicle handling profiles. Each car has its own acceleration, braking, grip, and turn response values — so the F1 car feels snappy and twitchy while the SUV feels heavy and sluggish. Collisions with traffic cars trigger a knock-back system that launches them with randomised rotation velocities and a basic bounce on landing. Building collisions just reverse your momentum so you bounce off rather than clip through. Of course I do want to improve the model, try and make it a bit more realistic, if I can I will try and model a ‘damaged’ version of the car that the model shifts to when involved in a collision.
Got the base city sim running in the game with a procedural infinite chunk system, day/night cycle, and arcade crash physics. Started porting the Kenney City Kit (Commercial) pack to replace the placeholder box buildings — all 35 GLB models are registered and loading asynchronously with box fallbacks while they stream in. Ran into a few path issues getting the assets wired up correctly across the project folder structure, which ate some time. Also squashed a reverse gear bug caused by a double-negative in the physics loop.
There were some problems I encountered on the way while writing the code for times and setting changes, during development, daytime driving ran at a flawless 60 FPS. But the moment the clock hit 7:00 PM, performance completely tanked down to 12 FPS.I expected it to be the physics or memory leaks to be the culprit, but it was just a bad math problem with rendering shadows. I also learnt how to add different forms of lighting in the form of both lamp posts and head lights on the cars themselves.
Next up: Traffic Lights possibly?? Maybe improve the UI a bit more and also make sure the AI is behaving correctly. I also want to implement a GTA style police system, and possibly a currency system where you earn credits when u crash into cars which can later be spent on getting better car models. Any suggestions, please let me know!
Dev Log 5-
Spent a good few hours setting up a much cleaner, more responsive user interface. This included building out a dedicated garage control panel that features an interactive headlight switch synced perfectly with the existing keyboard bindings. The HUD and telemetry gauges were also refined to give better layout feedback on speed and active vehicle statuses without cluttering the screen.The structural layout of the world received a major upgrade as well.
The building generation logic was completely reworked to create a significantly better, more immersive cityscape. Spacing, height varieties, and blocks were tuned alongside an updated day/night cycle that properly adjusts atmospheric fog and ambient lighting based on the position of the sun.Additionally, a massive chunk of time went into getting the AI traffic to actually work correctly. The tracking matrices were refactored so that the procedural cars recognize both the player and fellow commuters, allowing them to form clean lanes and slow down rather than creating clipping pileups.
On top of that, the crash physics were successfully re-integrated; slamming into a target now calculates a kinetic knock-back vector, sending the AI spinning and launching into a gravity-bound arc that provides maximum impact feedback while keeping the chaos contained. Next up on the list will be implementing functional traffic light cycles at the intersections and applying building textures to phase out the flat solid colours.
Dev Log 4 - 3D car game.
The project has moved beyond a simple driving prototype into a more advanced 3D traffic simulation. The system now supports imported car models, a working camera control system, and improved AI traffic behaviour. The focus has shifted from basic movement to building a more structured traffic environment with multiple systems interacting at once.
Vehicle System
The vehicle system has been upgraded from simple placeholder meshes to fully imported 3D car models. This improves visual quality and makes traffic feel more varied and realistic.
AI traffic now uses a random selection system when spawning cars, which assigns different car models to different vehicles. This creates more visual variety on the road and makes traffic easier to read.
The player vehicle is still handled separately from AI traffic, which keeps control responsive and prevents interference between systems.
AI Traffic System
The AI system currently uses lane-based movement where cars follow grid-aligned directions. Each car is assigned a direction on spawn and continues moving along that axis.
Traffic spawning has been improved so cars appear in structured positions aligned with the road grid rather than random world coordinates.
Basic interaction between cars is also present, including collision checks and spacing logic to prevent overlap.
However, there is currently a major issue where traffic sometimes stops moving completely. This is likely caused by cars entering a blocked state that is not properly reset or a velocity value not being restored after collision handling. This is the next major bug to fix.
Camera System
A mouse-controlled camera panning system has been added. This allows the player to look around while driving instead of being locked to a fixed camera angle.
The camera still follows the player smoothly but now includes manual input, making movement feel more natural and giving better spatial awareness in traffic.
World Structure
The environment is built on a grid-based road system. Roads are aligned along fixed axes, which provides structure for both player movement and AI navigation.
This grid system allows for predictable traffic behaviour and supports scaling the world without major changes to core logic.
The world is still visually simple, but the underlying structure is stable and supports further expansion such as intersections and traffic rules.
Known Issues
Traffic occasionally stops moving entirely and does not resume. This is the most important issue currently affecting the simulation.
Some AI cars become stuck at intersections due to collision or state handling not resetting correctly.
There may also be mismatches between car rotation and movement direction in some cases, especially after collisions or model changes.
These issues suggest that the AI system currently lacks a proper state management structure.
Next Steps
The next priority is to fix the traffic freeze bug by reviewing how movement states and collision responses are handled.
After that, the AI system should be refactored into a proper state-based system. Each car should have clear states such as moving, waiting, and blocked, rather than relying on simple speed changes.
Once stability is improved, the next major upgrade will be intersection logic. This will include controlled turning behaviour and basic traffic rules such as right of way or traffic lights.
A longer-term improvement will be replacing random movement with destination-based navigation so cars actually travel from point A to point B instead of spawning and moving endlessly.
Dev Log 3 – 3D Car Game (Traffic + World System Update)
Today the focus was on making the world feel less empty and giving AI cars more believable behaviour at junctions.
The environment now uses a grid-based road system, where roads run in straight lines across both axes. AI cars are spawned directly on these roads and are assigned a fixed direction along either the X or Z axis. This keeps traffic aligned with the lanes instead of drifting onto the grass.
A chunk-based building system was added so the world is no longer limited to a fixed square. Buildings are generated around the player as they move, based on which chunk of the map they are currently in. This makes the environment feel continuous and removes the visible edge of the map.
Intersections are now detected using the road grid. When a car reaches a point where both the X and Z positions align with a grid crossing, it is treated as an intersection. This allows the game to recognise junctions dynamically instead of hardcoding them.
A basic intersection blocking system was also introduced. Cars now check whether another vehicle is already using the same intersection. If it is occupied, the car will stop before entering and wait until it becomes clear. This prevents multiple cars from crossing at the same time and reduces overlap at junctions.
Traffic behaviour was adjusted so cars no longer rely on simple frame skipping. Instead, they actively pause movement when blocked and resume once the intersection is free. This makes the stopping behaviour more visible and consistent during gameplay.
Several bugs were fixed during this update, including incorrect chunk key formatting, a broken loop structure in the chunk generation system, a typo in traffic positioning, and inconsistencies in how cars were rotated based on direction. The intersection detection range was also widened slightly to make triggering more reliable.
There are still limitations in the current system. Cars do not yet form proper queues when waiting, and they do not make turning decisions at intersections. The blocking logic is also very simple, allowing only one car through at a time without priority rules or traffic lights.
Future improvements will focus on making traffic flow more natural. This includes adding proper queuing behaviour, introducing traffic lights, allowing cars to turn at intersections, and improving braking so cars slow down gradually instead of stopping instantly.
Added Roads and Building refinement - fixed the original AI traffic movement from aimlessly moving around, now they follow lanes properly (however some travel sideways in a certain direction of road layout which I plan to fix shortly)
Dev Log - 3D Car City Game
What got built
⸻
Key systems working now
⸻
Problems found and fixed
⸻
Next planned upgrades
⸻
Current state
Playable prototype, but still early stage:
