LanxinL / deepLearnigInPdcOfAutoDrivingPaperList

Paper list for deep learning in deciding and planning of auto-driving

deepLearnigInPdcOfAutoDrivingPaperList

Paper list for deep learning in deciding and planning of auto-driving

Planning

Interpretation

• 2020 Interpretable End-to-end Urban Autonomous Driving with Latent Deep Reinforcement Learning
  • 用 PGM 引入可解释的 latent Model
  • 用 Rl train PGM

Tuning

• Waymo (July 2019) & DeepMind - PBT (NOV 2017)
  • To make this process more efficient, researchers at DeepMind devised a way to automatically determine good hyperparameter schedules based on evolutionary competition (called “Population Based Training” or PBT), which combines the advantages of hand-tuning and random search.
  • PBT 的思路是，在training的每一步中都评分一下参数，然后选择新参数；
    • 这样会比train完再评价+选新参数效率更高，而且是在训练过程中的搜索（PBT discovers a schedule of hyperparameter settings rather than following the generally sub-optimal strategy of trying to find a single fixed set to use for the whole course of training.）
• Apollo-2018 An Auto-tuning Framework for Autonomous Vehicles
  • 将 AC 架构下的 IRL 简化至 Value Based
  • 假设human的得分应该在最高分附近，因此通过约束planner的得分要高于human的得分，训练cost func里的参数

Safety

• 2020Training Adversarial Agents to Exploit Weaknesses in Deep Control Policies用对抗agent去寻找当前pdc的弱点
  • 文章没写清楚input，output和model，缺少关键信息
• 2018 Parallel Planning: A New Motion Planning Framework for Autonomous Driving
  • train一个紧急情况的预测器，再train一个在紧急情况下直接生成trajectory的end-to-end的policy，以避免紧急情况下链路太长导致的相应不及时
  • 私认为该**在高速公路上比较适用，在公路和末端train复杂的policy无太大意义，但简化的policy会有用

Attention

• 2020 Learning a Directional Soft Lane Affordance Model for Road Scenes Using Self-Supervision
  • A self-supervised method for training a probabilistic network model to estimate the regions humans are most likely to drive in as well as a multimodal representation of the inferred direction of travel at each point.
• 2019-nips Social Attention for Autonomous Decision-Making in Dense Traffic
  • 用 attention 解决了周围障碍物排序的问题

Replace Cost Function

• CVPR-2019 End-to-end Interpretable Neural Motion Planner 新的 cost map 的建模方式
  • Input and Output
    • A holistic model that takes as input raw LIDAR data and a HD map and produces interpretable intermediate representations in the form of 3D detections and their future motion forecasted over the planning horizon.
    • Our final output representation is a space-time cost volume that represents the “goodness” of each possible location that the SDV can take within the planning horizon.
    • Our planner then scores a series of trajectory proposals using the learned cost volume and chooses the one with the minimum cost.
  • Loss function
    • 
    • Our planning loss encourages the minimum cost plan to be similar to the trajectory performed by human demonstrators.Note that this loss is sparse as a ground-truth trajectory only occupies small portion of the space. As a consequence, learning with this loss alone is slow and difficult.
    • To mitigate this problem, we introduce an another perception loss that encourages the intermediate representations to produce accurate 3D detections and motion forecasting.This ensures the interpretability of the intermediate representations and enables much faster learning.
• 2019 Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles
  • An interpretable cost function on top of perception, prediction and vehicle dynamics, and a joint learning algorithm that learns a shared cost function employed by our behavior and trajectory components.

End to end

via Imitation Learning

• Google brain & Waymo ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst
  • RNN
  • Input: consists of several images of size W*H pixels rendered into this top-down coordinate system.（sec 3.1）
  • Output: a drivable trajectory.

via Condition Imitation Learning

• ICCV-2019 Exploring the Limitations of Behavior Cloning for Autonomous Driving
  • policy模型采用ResNet，与预测模型共享特征提取层，以期望提取到更好的信息，增加了一些关于如何判断done等细节设计
  • 分析了目前Imitation Learning的不足（sec 5.3）
    • Generalization in the presence of dynamic objects
    • Driving Dataset Biases
    • Causal confusion and the inertia problem
    • High Variance

Control

• Path Tracking Control Using Imitation Learning with Variational Auto-Encoder ICCAS(rank - B4) 2019
• Trajectory VAE for multi-modal imitation ICLR(rejected) 2019
• Variational Autoencoder for End-to-End Control of Autonomous Driving with Novelty Detection and Training De-biasing IROS 2018

Simulator

• 2019 Game-Theoretic Modeling of Multi-Vehicle Interactions at Uncontrolled Intersections
  • a framework based on game theory for modeling vehicle interactions at uncontrolled intersections.
  • the model exhibits reasonable behavior expected in traffic, including the capability of reproducing scenarios extracted from real-world traffic data and reasonable performance in resolving traffic conflicts.