13 posts tagged with "Autonomous Driving"

CARLA ships an experimental Digital Twin Tool that turns a region of OpenStreetMap data into a procedurally generated 3D driving environment — road network, buildings, and all — directly inside the Unreal Engine editor. I wanted to try it on a small Plymouth region. What I actually got was a multi-day debugging session across a C++ renderer, a Blueprint-facing plugin, and the Unreal Editor itself before the pipeline would even run end-to-end. This post is about that debugging process, not a victory lap — the generated map still has real problems, which I'll get into at the end.

Simulators are great for autonomous driving testing because they're safe, reproducible, and fully controllable — but their rendered worlds usually look nothing like the real world. A perception or end-to-end driving model that performs well in simulation can behave very differently once it sees real camera data. This is the classic sim-to-real gap, and it's the problem a recent paper called Cam2Sim: Neural Scenario Reconstruction for Closed-Loop Autonomous Driving Simulation sets out to close.

If you've spent any time in CARLA, you know its strength and its limit at the same time: it's a fantastic, fully controllable simulator, but its Unreal Engine art assets look like a simulator. Buildings, road textures, foliage — it's all clearly synthetic. For a lot of autonomous driving development, that's fine. But when you're trying to close the sim-to-real gap, that visual gap matters.

I recently deployed RF-DETR — Roboflow's real-time, transformer-based object detector — into a TensorRT-accelerated ROS2 pipeline. This post covers what RF-DETR is, why I picked it, and how the deployment is structured.

After my previous experiments with camera-based detection and FLOAM-based LiDAR odometry in CARLA, I wanted to push things further — this time focusing on robustness under adversarial conditions.

After my previous CARLA experiment with camera-based object detection, I wanted to push further — this time focusing on LiDAR and localization.

I recently started experimenting with the CARLA simulator and decided to test whether a pre-trained perception model — with zero exposure to simulated data — could hold up against synthetic imagery.

After completing my SCNN lane detection project, I reviewed the RESA paper (2020), which extends SCNN's message-passing mechanism with a Recurrent Feature Shift Aggregator (RESA) and a Bilateral Up-Sampling Decoder (BUSD). The paper claimed a 9% F1-score improvement over SCNN. My experiments told a different story.

I recently worked on a lane marking detection project. After reviewing the literature, I decided to use the SCNN (Spatial CNN) model — not because it's state of the art, but because it falls squarely into my "if you can't implement it, you don't really understand it" category.

I recently deployed three deep learning models covering three core perception tasks — all running simultaneously in real-time on a single GPU.