Core Viewpoint - The article presents a framework for generating safe trajectories based on discrete diffusion without the need for gradient calculations, aiming to enhance the efficiency and safety of autonomous driving systems [1][2]. Group 1: Framework Overview - The core value lies in introducing discrete ideas into trajectory generation for assisted driving, moving beyond traditional reinforcement learning and continuous diffusion models [2]. - The framework consists of two main phases: goal-oriented trajectory generation and safety-guided regeneration, both of which do not require gradient calculations [5][10]. Group 2: Phase One - Goal-Conditioned Generation - The objective is to generate a diverse set of complete trajectory plans reflecting different high-level driving intentions, crucial for scenarios requiring extensive decision-making [3]. - The workflow includes generating candidate target points, ensuring diversity through non-maximum suppression (NMS), generating complete trajectories for each candidate, and selecting the best trajectory based on a global scoring system [4]. Group 3: Phase Two - Safety-Guided Regeneration - This phase focuses on iteratively correcting trajectories identified as potentially unsafe through a dialogue between the generated model and an external safety oracle [5][10]. - The process involves trajectory evaluation, safety anchor search, and trajectory inpainting to ensure the final trajectory is safe and coherent [7][8][9]. Group 4: Challenges and Innovations - The article highlights challenges in relying solely on reinforcement learning due to issues like reward hacking and the difficulty of ensuring safety in imitation learning [11]. - It emphasizes the need for a unified architecture that efficiently integrates multimodal inputs while maintaining safety and performance, leading to the development of the ReflectDrive framework [13].

Core Viewpoint - ReflectDrive is a novel learning framework that integrates a reflective mechanism to achieve safe trajectory generation through discrete diffusion, addressing the challenges in end-to-end autonomous driving systems [4][46]. Group 1: Introduction and Background - Autonomous driving is leading the transportation industry towards a safer and more efficient future, with end-to-end (E2E) systems becoming a mainstream alternative to traditional modular designs [4]. - Visual-Language-Action (VLA) models combine pre-trained knowledge from visual-language models (VLM) to enhance adaptability in complex scenarios [4][5]. - Current learning-based methods have not resolved core challenges in imitation learning driving systems, particularly in encoding physical rules like collision avoidance [4][5]. Group 2: ReflectDrive Framework - ReflectDrive proposes a new learning framework that utilizes a discrete diffusion reflective mechanism for safe trajectory generation [3][12]. - The framework begins by discretizing the two-dimensional driving space to construct an action codebook, allowing fine-tuning of pre-trained diffusion language models for planning tasks [3][14]. - The reflective mechanism operates without gradient calculations, enabling iterative self-correction inspired by spatiotemporal joint planning [3][8]. Group 3: Methodology and Mechanism - The reflective inference process consists of two stages: target condition trajectory generation and safety-guided regeneration [20][25]. - The framework integrates safety metrics to evaluate generated multimodal trajectories, identifying unsafe path points through local search methods [8][25]. - The iterative optimization loop continues until the trajectory is deemed safe or computational limits are reached, ensuring high efficiency in real-time performance [31][32]. Group 4: Experimental Results - ReflectDrive was evaluated on the NAVSIM benchmark, demonstrating significant improvements in safety metrics such as collision rates and compliance with drivable areas [32][38]. - The introduction of the safety-guided regeneration mechanism led to substantial enhancements in safety indicators, with notable increases in DAC (3.9%), TTC (1.3%), NC (0.8%), and EP (7.9%) compared to the baseline [37][38]. - When using ground-truth agent information, ReflectDrive's performance approached human driving levels, achieving NC of 99.7% and DAC of 99.5% [38][39]. Group 5: Conclusion - ReflectDrive effectively integrates a reflective mechanism with discrete diffusion for safe trajectory generation, validated by its performance on the NAVSIM benchmark [46].