[Paper Review] AutoDRIVE -- Technical Report

Tanmay Vilas Samak, Chinmay Vilas Samak|arXiv (Cornell University)|Nov 15, 2022

Transportation and Mobility Innovations4 citations

TL;DR

AutoDRIVE is an open-source, integrated cyber-physical platform for autonomous driving research and education, comprising a 1:14 scale testbed vehicle, a ROS-based simulator, and a Devkit for algorithm development. It enables end-to-end implementation of autonomy algorithms—including autonomous parking, behavioral cloning, intersection traversal, and smart city management—using real hardware, simulation, and a unified software stack with support for single and multi-agent systems.

ABSTRACT

This work presents AutoDRIVE, a comprehensive research and education platform for implementing and validating intelligent transportation algorithms pertaining to vehicular autonomy as well as smart city management. It is an openly accessible platform featuring a 1:14 scale car with realistic drive and steering actuators, redundant sensing modalities, high-performance computational resources, and standard vehicular lighting system. Additionally, the platform also offers a range of modules for rapid design and development of the infrastructure. The AutoDRIVE platform encompasses Devkit, Simulator and Testbed, a harmonious trio to develop, simulate and deploy autonomy algorithms. It is compatible with a variety of software development packages, and supports single as well as multi-agent paradigms through local and distributed computing. AutoDRIVE is a product-level implementation, with a vast scope for commercialization. This versatile platform has numerous applications, and they are bound to keep increasing as new features are added. This work demonstrates four such applications including autonomous parking, behavioural cloning, intersection traversal and smart city management, each exploiting distinct features of the platform.

Motivation & Objective

• To develop a comprehensive, accessible platform for research and education in autonomous driving and smart city systems.
• To bridge the gap between simulation and real-world deployment through a unified hardware-software ecosystem.
• To support diverse autonomy paradigms, including model-based and data-driven approaches, across single and multi-agent scenarios.
• To demonstrate the platform’s capabilities through real-world and simulated implementations of key autonomous driving applications.
• To enable scalable, extensible, and commercially viable research infrastructure for academia and industry.

Proposed method

• Design and fabrication of a 1:14 scale autonomous vehicle (Nigel) with redundant sensing, actuation, and high-performance computing.
• Development of a ROS-based simulator with support for vehicle dynamics, infrastructure modeling, and configurable graphics and lighting.
• Creation of the AutoDRIVE Devkit, a Python API-enabled interface for seamless algorithm deployment to the testbed.
• Implementation of modular software stacks for autonomous driving and smart city management, enabling integration with ROS and web-based control.
• Use of SLAM, probabilistic localization, and end-to-end learning pipelines (e.g., behavioral cloning) for navigation and control.
• Support for distributed and local computing, enabling both single-vehicle and multi-vehicle coordination in simulation and real-world testing.

Experimental results

Research questions

• RQ1 How can a unified platform integrate hardware prototyping, simulation, and software development for autonomous driving research?
• RQ2 To what extent can a scaled-down testbed replicate real-world autonomous driving challenges in a controlled, accessible environment?
• RQ3 How effective is the AutoDRIVE platform in supporting diverse autonomy paradigms, including end-to-end learning and modular control?
• RQ4 What is the performance of the platform in enabling sim-to-real transfer for behavioral cloning and autonomous navigation?
• RQ5 How scalable and extensible is the platform for future integration of heterogeneous vehicles, robotic pedestrians, and large-scale urban scenarios?

Key findings

• AutoDRIVE successfully demonstrated autonomous parking using SLAM and probabilistic localization, achieving stable navigation in a predefined map.
• Behavioral cloning was implemented with data augmentation and sim2real transfer, enabling the vehicle to mimic human driving behavior on the testbed.
• The platform enabled robust intersection traversal using a centralized control architecture, with real-time response to traffic signals and signs.
• Smart city management was validated through a centralized SCM server that coordinated vehicle behavior with traffic light control and surveillance.
• The platform supports multi-agent scenarios through distributed computing and enables seamless deployment from simulation to real hardware via the Devkit API.
• The system’s modular design and open architecture allow for extensibility, with potential for integration of heterogeneous vehicles and advanced urban mobility applications.