EDGAR: An Autonomous Driving Research Platform -- From Feature Development to Real-World Application (2309.15492v2)

Abstract: While current research and development of autonomous driving primarily focuses on developing new features and algorithms, the transfer from isolated software components into an entire software stack has been covered sparsely. Besides that, due to the complexity of autonomous software stacks and public road traffic, the optimal validation of entire stacks is an open research problem. Our paper targets these two aspects. We present our autonomous research vehicle EDGAR and its digital twin, a detailed virtual duplication of the vehicle. While the vehicle's setup is closely related to the state of the art, its virtual duplication is a valuable contribution as it is crucial for a consistent validation process from simulation to real-world tests. In addition, different development teams can work with the same model, making integration and testing of the software stacks much easier, significantly accelerating the development process. The real and virtual vehicles are embedded in a comprehensive development environment, which is also introduced. All parameters of the digital twin are provided open-source at https://github.com/TUMFTM/edgar_digital_twin.

• The paper introduces EDGAR’s comprehensive system architecture that integrates multi-sensor data and simulation-based testing for real-world validation.
• The paper outlines a high-fidelity digital twin methodology that synchronizes sensor inputs to ensure consistent performance assessments in autonomous vehicles.
• The paper emphasizes an iterative workflow that bridges the simulation-to-reality gap, enhancing feature development and deployment in urban driving.

Overview of "EDGAR: An Autonomous Driving Research Platform -- From Feature Development to Real-World Application"

The paper presents the development and deployment of EDGAR, an autonomous driving research platform established to facilitate the transition from feature development to real-world application in autonomous vehicles (AVs). The authors tackle the complexities inherent in the integration, validation, and performance assessment of broad-ranging autonomous software stacks. Emphasizing simulation-based testing, the paper introduces EDGAR's digital twin, enhancing methodological consistency across virtual and real-world tests.

EDGAR stands as the culmination of an intricate integration of hardware and software design facets. The platform is characterized by a multi-sensor setup using cameras, LiDARs, RADARs, and microphones, backed by dual high-performance computing systems and network projects capable of handling large volumes of data—an ambitious feat designed to address perception and planning. The digital twin is publicly accessible, emphasizing vehicle dynamics and sensor fidelity. Critical to its deployment functionality is the standardized use of the Precision Time Protocol (PTP) to synchronize sensor timestamps in real-time.

Core Contributions and Experimental Insights

1. System Architecture: The architecture of EDGAR emphasizes redundancy and scalability, with flexible modular interfaces adaptable for diverse research themes. This flexibility aids in the evaluation of varying sensor configurations and computational strategies facilitating interdisciplinary research between domains such as teleoperations and human-machine interfaces (HMI).
2. Digital Twin: The digital twin components encompass vehicle dynamics models alongside sensor and network replication. This fidelity supports consistent performance assessments from simulation to reality, which is paramount for developing validation methodologies emphasizing comprehensive and reproducible testing strategies for AVs.
3. Holistic Workflow: The workflow integrates feature development with validation phases that include Model-in-the-Loop, Software-in-the-Loop, Hardware-in-the-Loop, and real-world testing. This structured pipeline allows iterative development with feedback loops, contributing valuable real-world data back into simulation environments, thus addressing the Sim2Real gap.
4. Research Implications: The paper’s insights pertain directly to enhancing autonomy in urban driving contexts, with EDGAR's research platform supporting a wide breadth of studies on perception, planning, and control. By making all parameters of the digital twin open-source, it invites the scientific community to participate, further validating and broadening this research initiative.

Evaluation and Future Prospects

The paper rigorously examines the gap between simulation and physical deployments by juxtaposing the digital twin against live vehicle data. Observations on calibration and synchronization processes enhance understanding, albeit highlighting discrepancies between ideal and operational conditions. Improvements in sensor modeling and triggering mechanisms are suggested pathways to bridge the remaining Sim2Real gap.

Future directions involve refining digital twin accuracy through the reflection of recorded anomalies and expanding the dataset with complex urban settings. The platform's ability to adapt to real-world challenges promises incremental advancements in AI research, particularly in encoding edge case scenarios and integrating machine learning methodologies focused on safety and efficacy.

In conclusion, EDGAR is exemplary in demonstrating a comprehensive system executing the full cycle from development to deployment. It provides crucial contributions to AV technology, laying a robust foundation for advancing autonomous vehicles to navigate public road scenarios reliably. The open dissemination of its digital twin establishes a collaborative milestone for the community, setting a benchmark for consistency and reproducibility in AV research.