AI-CDA4All: Democratizing Cooperative Autonomous Driving for All Drivers via Affordable Dash-cam Hardware and Open-source AI Software

Shengming Yuan, Hao Zhou* · 2025

Read AI-CDA4All: Democratizing Cooperative Autonomous Driving for All Drivers via Affordable Dash-cam Hardware and Open-source AI Software by Shengming Yuan, Hao Zhou* (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-06749.txt.

Introduction

Yuan and Zhou present a 2025 paper on bringing cooperative self-driving to ordinary cars through cheap dash-cams and open software. Connected-vehicle systems usually demand pricey sensors and units that lock into one brand. Their kit leans on mass-market edge devices drivers already can buy. Light vehicle-to-everything links run over LTE and WiFi instead of dedicated radios.

Standards alignment keeps the kit compatible with planned infrastructure. The team targets public gains first, such as smoother traffic flow management at scale. Feasibility tests check whether the low-cost links carry data fast enough for cooperation. The pitch is autonomy's safety benefits without a luxury price tag or a new car.

Catalog record

CREATORShengming Yuan, Hao Zhou*
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-06749

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

AI-CDA4All: Democratizing Cooperative Autonomous Driving for All Drivers via Affordable Dash-cam Hardware and Open-source AI Software Shengming Yuan, Hao Zhou* Abstract As transportation technology advances, the demand for connected vehicle infrastructure has greatly increased to improve their efficiency and safety. One area of advancement, Cooperative Driving Automation (CDA) still relies on expensive autonomy sensors or connectivity units and are not interoperable across existing market car makes/models, limiting its scalability on public roads. To fill these gaps, this paper presents a novel approach to democratizing CDA technology, it leverages low-cost, commercially available edge devices such as vehicle dash-cams and open-source software to make the technology accessible and scalable to be used in transportation infrastructure and broader public domains. This study also investigates the feasibility of utilizing cost-effective communication protocols based on LTE and WiFi. These technologies enable lightweight Vehicle-to-Everything (V2X) communications, facilitating real-time data exchange between vehicles and infrastructure. Our research and development efforts are aligned with industrial standards to ensure compatibility and future integration into existing transportation ecosystems. By prioritizing infrastructure-oriented applications, such as improved traffic flow management, this approach seeks to deliver tangible societal benefits without directly competing with vehicle OEMs. As recent advancement of Generative AI (GenAI), there is no standardized integration of GenAI technologies into open-source CDAs, as the current trends of muiltimodal large language models gain popularity, we demonstrated a feasible locally deployed edge LLM models can enhance driving experience while preserving privacy and security compared to cloud-connected solutions. The proposed system underscores the potential of low-cost, scalable solutions in advancing CDA functionality, paving the way for smarter, safer, and more inclusive transportation networks. Fig. 1: AI-CDA4ALL Ecosystem I Introduction Autonomous driving technologies have seen rapid advancements in recent years, yet they remain largely inaccessible to the general public. Most commercial systems are built on expensive sensor suites, proprietary algorithms, and closed vehicle platforms, making them cost-prohibitive and non-interoperable across different car makes and models. This is particularly limiting given that, since 2016, the majority of production vehicles have adopted electronic controlled, digitally manage core driving functions such as steering, braking, and acceleration. Through community-driven reverse engineering, these control interfaces have been unlocked across hundreds of car models [ 1 ] , revealing that low-cost, software-based autonomy is technically feasible using readily available consumer hardware—such as a smartphone or dash-mounted camera—as the primary computation and sensing platform. In parallel, open-source self-driving frameworks such as OpenPilot [ 2 ] , Autoware [ 3 ] , and others have significantly lowered the barrier to autonomous development. These platforms enable individuals and researchers to design and test perception, planning, and control modules on real vehicles, offering high customizability and extensibility. Furthermore, they present a powerful opportunity for infrastructure stakeholders—such as cities or transportation agencies—to build advisory or cooperative driving applications directly into the vehicle software stack. While community developers have embraced these systems for personal projects, the research community and public infrastructure stakeholders have yet to fully engage. Many academic innovations remain confined to in-house testbeds and fail to transition into scalable, real-world deployments. Infrastructure owners, in turn, face barriers in interfacing with proprietary OEM systems. This disconnect creates a bottleneck between research outcomes and real-world impact. We argue that a more open and participatory ecosystem is needed—one that empowers customers, researchers, and infrastructure stakeholders to collaboratively build and improve connected autonomy systems. A critical component of cooperative driving is robust vehicle connectivity. Yet most high-performance communication solutions—such as PC5 [ 4 ] used in advanced V2X solutions, require dedicated and expensive onboard units, severely limiting scalability. Cellular and WiFi-based alternatives have historically been dismissed due to performance limitations, but the landscape has changed: modern protocols like LTE-Advanced, 5G, and WiFi 6 offer significantly improved reliability and throughput [ 5 ] . These advances make them promising candidates for non-safety-critical V2I and V2X communications, such as traffic advisory messages, infrastructure updates, and digital twin interactions. In this work, we revisit the viability of these modern wireless protocols for connected driving, evaluating their performance and integrating them into the open-source self-driving stack. To support broader use, we provide standardized APIs for connectivity-based functions and enable infrastructure applications to interface with vehicles directly, reducing deployment friction. Simultaneously, the rapid emergence of generative AI (GenAI) models, particularly multimodal large language models, offers exciting potential to enhance driving autonomy through reasoning, personalization, and decision support. Despite early explorations in the literature [ 6 ] , GenAI integration into real-time driving and connectivity systems remains limited. A key obstacle is the absence of standardized interfaces that link GenAI capabilities with widely used open-source self-driving stacks. To address this, we build a lightweight API layer that connects GenAI models with both driving logic and communication modules. In addition, we introduce meta-action functions based on the internal design of the baseline self-driving system, allowing GenAI models to translate abstract reasoning into executable commands more accurately and efficiently. In this paper , we aim to address these critical research gaps by building an open-source, affordable, and standardized CDA system with integrated AI capabilities. Our goal is to democratize AI-enabled CDA across three key communities: everyday drivers, academic researchers, and infrastructure stakeholders. By enabling scalable real-world deployments, fostering interoperability, and supporting intelligent interactions between vehicles and infrastructure, we seek to accelerate the development of a more open, inclusive, and collaborative ecosystem for connected and autonomous transportation. Motivated by the aforementioned research gap and guided by a transportation engineering perspective , this paper contributes to the following main efforts: 1. Scalable and Affordable Open-source Autonomy : We build upon a widely adopted Level-2 open-source self-driving platform to ensure affordability, extensive compatibility with existing car models, and a straightforward plug-and-play experience. We have tested our platform extensively on over 10 different car make and model. 2. Lightweight Connectivity Development and Integration : We proposed, implemented and demonstrated a lightweight, standards-compliant connectivity system that adheres to SAE V2X standards J2735 [ 7 ] , which is able to transmit non-safety-critical ones (e.g., toll information) using low cost protocol over WiFi and LTE connections. Transmitting message such as advisory speed messages to alleviate congestion, enhance road safety, and enable cooperative driving (e.g., queue formation and merging assistance). 3. API Development for GenAI Integration : We design a suite of custom APIs (i.e. MetaAction) that bridge the gap between high-level LLM outputs and executable commands for scalable open-source autonomy. Through code-generation mechanisms, Large Language Models (LLMs) produce structured, valid code segments that can be parsed and executed, thus providing a robust method for incorporating GenAI-based decision-making into real on-road scenarios. We provided a foundational development ground for any future Generative AI platform to integrate to AI-CDA4ALL . II Related Work Accessibility of Current Autonomous Driving : Many existing research papers on Connected and Automated Vehicles (CAVs) have demonstrated their solutions on one or two specific car models, typically using proprietary hardware or manufacturer-specific control systems [ 8 , 9 ] . These implementations, while valuable as proofs of concept, lack scalability and interoperability, making them challenging to deploy across a wider range of vehicles. This pose a significant gap between research community and industry which both, does not use open source system and cannot transfer technology to each other. In contrast, our approach

Read the complete text