Publication ·
New preprint: Scale-CDA brings AI-assisted cooperative driving to production cars for under $1,000
A new lab preprint introduces Scale-CDA, an open-hardware/open-software toolchain that retrofits ordinary production cars for generative-AI-assisted cooperative driving automation using off-the-shelf parts costing under US $1,000.
A new preprint from the lab — Scale-CDA: a scalable prototype to democratize AI-assisted cooperative driving automation (CDA) for production cars, by Hao Zhou, Shengming Yuan, Yuhang Wang, and Haibin Wen — is now on arXiv.
Cooperative driving automation research has long been gated by cost. Field trials have needed purpose-built instrumented vehicles, which caps most studies at a handful of cars. Scale-CDA takes the opposite approach: build on the community-maintained OpenDBC interface (300+ car models) and Openpilot Level-2 ADAS, and retrofit an ordinary car with off-the-shelf parts — an edge PC, a webcam, a CAN adapter, and optional LTE/Wi-Fi radios — for under US $1,000.
On top of that base, the paper adds a lightweight MQTT-over-Wi-Fi-6/LTE V2X stack, an edge-deployed multimodal LLM that reads synchronized vision, CAN, and V2X streams through a Model-Context-Protocol bridge, and a library of meta-action executors that translate the model’s structured JSON advisories into verified Openpilot planner hooks — lane changes, gap management, emergency stops — without touching the safety-certified control core.
Field testing on a 7.5 km loop measured 5.25 ms mean round-trip latency and link speeds near 100 Mb/s; in multi-vehicle road tests the full stack held end-to-end decision latency below 60 ms, with all inference kept on-board so driving data never leaves the vehicle.
The bills-of-materials, connectivity APIs, and GenAI bridges will be released as open resources. More detail on the project page.