LimSim Series: An Autonomous Driving Simulation Platform for Validation and Enhancement

TL;DR

The paper tackles the challenges of validating and enhancing autonomous driving systems in closed-loop simulations by balancing realism, efficiency, and comprehensive evaluation. It presents LimSim Series, an open-source, modular platform that integrates multi-type road-network information, AoI-based resource optimization, and background-vehicle decision-making, accompanied by baseline algorithms, user-friendly interfaces, and multi-dimensional evaluation metrics. Through experiments across highway, ramp, intersection, roundabout, and long-route scenarios, the framework demonstrates compatibility with modular, end-to-end, and knowledge-driven ADS, enabling rapid deployment and iterative improvement. The authors also discuss design considerations, such as real-data integration and scalable simulation, and release code to accelerate community adoption. Future work outlines high-fidelity sensor rendering, heterogeneous traffic, and AI-driven corner-case generation to further improve realism and testing coverage.

Abstract

Closed-loop simulation environments play a crucial role in the validation and enhancement of autonomous driving systems (ADS). However, certain challenges warrant significant attention, including balancing simulation accuracy with duration, reconciling functionality with practicality, and establishing comprehensive evaluation mechanisms. This paper addresses these challenges by introducing the LimSim Series, a comprehensive simulation platform designed to support the rapid deployment and efficient iteration of ADS. The LimSim Series integrates multi-type information from road networks, employs human-like decision-making and planning algorithms for background vehicles, and introduces the concept of the Area of Interest (AoI) to optimize computational resources. The platform offers a variety of baseline algorithms and user-friendly interfaces, facilitating flexible validation of multiple technical pipelines. Additionally, the LimSim Series incorporates multi-dimensional evaluation metrics, delivering thorough insights into system performance, thus enabling researchers to promptly identify issues for further improvements. Experiments demonstrate that the LimSim Series is compatible with modular, end-to-end, and VLM-based knowledge-driven systems. It can assist in the iteration and updating of ADS by evaluating performance across various scenarios. The code of the LimSim Series is released at: https://github.com/PJLab-ADG/LimSim.

Figures (6)

• Figure 1: The LimSim Series is all you need for agile and effective autonomous driving simulation.
• Figure 2: Multiple technology pipelines for autonomous driving systems.
• Figure 3: (1) Driving Engine emphasizes independence and compatibility, allowing for full deployment and testing of algorithms while interfacing with open-source engines like SUMO and CARLA for cross-platform development. (2) Map construction involves importing OpenDrive-formatted files or obtaining data through cross-platform communication, utilizing Cartesian and Frenet coordinate systems for accurate vehicle positioning and trajectory planning. (3) Scene understanding is enhanced through 3D scene information from CARLA, enabling the use of sensor data for traffic participant identification and motion state estimation, which can be applied to perception algorithms or integrated into end-to-end architectures. (4)Decision and planning include traditional models and advanced joint decision planning models that leverage MCTS for behavior decisions and parallel trajectory planning for trajectory generation. (5) Performance evaluation is facilitated through a series of indicators focusing on vehicle operation status, with the ability to calibrate model parameters to align with real-world data for enhanced simulation realism.
• Figure 4: Differentiated decision-making planning strategies for other vehicles inside and outside the AoI.
• Figure 5: Interactive simulation strategy combines virtual simulation and real traffic data.