Summary of Gpd-1: Generative Pre-training For Driving, by Zixun Xie et al.
GPD-1: Generative Pre-training for Driving
by Zixun Xie, Sicheng Zuo, Wenzhao Zheng, Yunpeng Zhang, Dalong Du, Jie Zhou, Jiwen Lu, Shanghang Zhang
First submitted to arxiv on: 11 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a unified Generative Pre-training for Driving (GPD-1) model that can accomplish various tasks related to autonomous driving, including scene generation, traffic simulation, closed-loop simulation, map prediction, and motion planning. The GPD-1 model is based on an autoregressive transformer architecture with a scene-level attention mask, which enables intra-scene bi-directional interactions. The model also includes hierarchical positional tokenizers for ego and agent tokens, as well as a map vector-quantized autoencoder for compressing ego-centric semantic maps into discrete tokens. The GPD-1 is pre-trained on the nuPlan dataset and can generalize to various tasks without fine-tuning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper presents a new approach to modeling driving scenarios that combines multiple tasks, including scene generation, traffic simulation, closed-loop simulation, map prediction, and motion planning. This unified model, called GPD-1, uses a transformer architecture with attention mechanisms to generate scenes, predict maps, and plan motions. |
Keywords
» Artificial intelligence » Attention » Autoencoder » Autoregressive » Fine tuning » Mask » Transformer
