自动驾驶motion相关任务

自动驾驶motion相关任务

• 轨迹预测

• • paper list
  • • VectorNet
    • MultiPath
    • MultiPath++

• 规划

• 预测和规划一体

• 场景/交通流仿真与泛化

• • waymo
  • waabi Traffic Modelling

轨迹预测

任务定义: 给定场景(地图信息)和前1s物体的运动轨迹,预测这些物体的未来的轨迹(可能输出几条轨迹)

paper list

VectorNet

VectorNet: Encoding HD Maps and Agent Dynamics from Vectorized Representation
https://arxiv.org/abs/2005.04259

MultiPath

MultiPath: Multiple Probabilistic Anchor Trajectory Hypotheses for Behavior Prediction
https://arxiv.org/abs/1910.05449

MultiPath++

MultiPath++: Efficient Information Fusion and Trajectory Aggregation for Behavior Prediction
https://arxiv.org/abs/2111.14973

规划

预测和规划一体

game former

场景/交通流仿真与泛化

Guided Conditional Diffusion for Controllable Traffic Simulation
提出了一种基于扩散模型的可控交通仿真方法。

MotionDiffuser: Controllable Multi-Agent Motion Prediction Using Diffusion

TrafficSim: Learning to Simulate Realistic Multi-Agent Behaviors

SceneGen: Learning to Generate Realistic Traffic Scenes
何生成现实交通场景，与扩散模型有相似的目标

TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Prediction
自动驾驶模拟和运动预测的世界模型

waymo

https://waymo.com/research/?ncr=

SceneDiffuser: Efficient and Controllable Driving Simulation Initialization and Rollout
Improving Agent Behaviors with RL Fine-tuning for Autonomous Driving
UniGen: Unified Modeling of Initial Agent States and Trajectories for Generating Autonomous Driving Scenarios
MotionDiffuser: Controllable Multi-Agent Motion Prediction using Diffusion

waabi Traffic Modelling

Learning to Drive via Asymmetric Self-Play
SceneControl: Diffusion for Controllable Traffic Scene Generation
Learning Realistic Traffic Agents in Closed-loop
MixSim: A Hierarchical Framework for Mixed Reality Traffic Simulation link
Rethinking Closed-loop Training for Autonomous Vehicles
Towards Scalable Coverage-Based Testing of Autonomous Vehicles

[3]:
[4]: https://zhuanlan.zhihu.com/p/707223043