Yihan Hu,Siqi Chai,Zhening Yang,Jingyu Qian,Kun Li,Wenxin Shao,Haichao Zhang,Wei Xu,Qiang Liu
Abstract:As autonomous driving systems being deployed to millions of vehicles, there is a pressing need of improving the system's scalability, safety and reducing the engineering cost. A realistic, scalable, and practical simulator of the driving world is highly desired. In this paper, we present an efficient solution based on generative models which learns the dynamics of the driving scenes. With this model, we can not only simulate the diverse futures of a given driving scenario but also generate a variety of driving scenarios conditioned on various prompts. Our innovative design allows the model to operate in both full-Autoregressive and partial-Autoregressive modes, significantly improving inference and training speed without sacrificing generative capability. This efficiency makes it ideal for being used as an online reactive environment for reinforcement learning, an evaluator for planning policies, and a high-fidelity simulator for testing. We evaluated our model against two real-world datasets: the Waymo motion dataset and the nuPlan dataset. On the simulation realism and scene generation benchmark, our model achieves the state-of-the-art performance. And in the planning benchmarks, our planner outperforms the prior arts. We conclude that the proposed generative model may serve as a foundation for a variety of motion planning tasks, including data generation, simulation, planning, and online training. Source code is public at <a class="link-external link-https" href="https://github.com/HorizonRobotics/GUMP/" rel="external noopener nofollow">this https URL</a>
What problem does this paper attempt to address?
The paper primarily aims to address several key issues in autonomous driving systems, including improving system scalability, safety, and reducing engineering costs. Specifically, the research team proposed an efficient solution—a generative model-based approach to learn the dynamics of driving scenarios. Through this model, not only can various future situations of a given driving scenario be simulated, but different driving scenarios can also be generated based on different prompts. The unique aspect of this model is that it can operate in both full-Autoregressive and partial-Autoregressive modes, significantly enhancing training and inference speed without sacrificing its generative capabilities.
The main contributions mentioned in the paper can be summarized as follows:
1. **Proposed a new generative model**: Named GUMP (Generative Unified Model for Motion Planning), it employs a simple key-value pair tokenizer design that efficiently handles driving scene data and supports efficient autoregressive structures. The model not only achieved state-of-the-art performance in simulation realism and scene generation benchmarks but also performed excellently in planning benchmarks.
2. **Extensively explored the model's applications**: The research team discussed in detail how GUMP can be used as a foundational model for a series of downstream tasks such as data generation, simulation, planning, and online training, and demonstrated that it can significantly enhance the functionality of these tasks.
3. **Built a unified framework**: This framework uses the generative model as a core component to develop a closed-loop training and evaluation system. According to the authors, this is the first time a unified framework has been used to address all learning-based motion planning tasks.
Experimental results show that GUMP achieved significant results on multiple public datasets, including but not limited to:
- **Scene generation**: Outperformed other methods on the Waymo Open Dataset, particularly in terms of speed and size metrics, with errors reduced by 42.1% and 26.6%, respectively.
- **World simulator**: Excelled in interactive, map-based metrics in the Waymo Sim Agents benchmark and achieved the lowest overall minimum average displacement error (minADE), indicating state-of-the-art performance in realism metrics.
- **Interactive planning**: Conducted open-loop and closed-loop experiments on the nuPlan dataset, where GUMP outperformed other methods in most metrics in the open-loop experiments, validating its effectiveness as a world simulator.
In summary, this paper addresses multiple challenges in the field of autonomous driving by proposing a novel generative model, GUMP, and demonstrates its powerful application potential in various tasks.