by 张启超 2020.3.2
Self-Driving Cars: A Survey, 2019, [paper] (https://arxiv.org/pdf/1901.04407.pdf)
A Survey of Motion Planning and Control Techniques for Self-Driving Urban Vehicles, 2016, paper
Planning and Decision-Making for Autonomous Vehicles, MIT Reviewer, 2018, paper
Deep Reinforcement Learning for Autonomous Driving: A Survey, 2020, paper
A Survey of Deep Learning Applications to Autonomous Vehicle Control, IEEE Transaction on ITS 2019, paper
Learning a Driving Simulator, 2016, Comma.ai paper, code
End to End Learning for Self-Driving Cars, 2016, NVIDIA paper, code
Generative Adversarial Imitation Learning, NIPS 2016, paper
Robust End-to-End Learning for Autonomous Vehicles, MIT, 2018 Master Thesis
End-to-end Driving via Conditional Imitation Learning, 2018, ICRA paper, code
ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst, 2018, Waymo paper, code
Self-Imitation Learning, ICML 2018, paper, code
Multi-Agent Generative Adversarial Imitation Learning, ICLR 2018, paper
Conditional affordance learning for driving in urban environments, CoRL, 2018, paper, code
Learning by cheating， 2019， paper, code
Imitation Learning from Imperfect Demonstration, ICML, 2019, paper, code
Exploring the Limitations of Behavior Cloning for Autonomous Driving, ICCV, 2019, paper
Multimodal End-to-End Autonomous Driving， 2019， paper
DEEP IMITATIVE MODELS FOR FLEXIBLE INFERENCE, PLANNING, AND CONTROL， 2020，ICLR，paper, model
Deep Imitation Learning for Autonomous Driving in Generic Urban Scenarios with Enhanced Safety, 2020 paper
Maximum Entropy Deep Inverse Reinforcement Learning, 2015, paper
Generative Adversarial Imitation Learning, NIPS, 2016, paper
Guided Cost Learning: Deep Inverse Optimal Control via Policy Optimization, ICML 2016, paper
A Connection between Generative Adversarial Networks, Inverse Reinforcement Learning, and Energy-Based Models, NIPS 2016, paper
InfoGAIL: Interpretable Imitation Learning from Visual Demonstrations, NIPS 2017, paper, code
Courteous Autonomous Cars, IROS, 2018, paper, blog
Probabilistic Prediction of Interactive Driving Behavior via Hierarchical Inverse Reinforcement Learning, ITS, 2018, paper, blog
Learning Robust Rewards with Adversarial Inverse Reinforcement Learning, ICLR 2018, paper
Multi-Agent Adversarial Inverse Reinforcement Learning, ICML 2019, paper
Learning Driving decisions by imitating drivers' control behaviors, 2019, [paper] (https://arxiv.org/pdf/1912.00191.pdf)
Towards Learning Multi-agent Negotiations via Self-Play, ICCV 2019, paper
CIRL: Controllable Imitative Reinforcement Learning for Vision-based Self-driving, ECCV 2018, paper, code
Imitating Driver Behavior with Generative Adversarial Networks, IV 2017, [paper] [code]
Multi-Agent Imitation Learning for Driving Simulation, IROS 2018, [paper] [code]
Simulating Emergent Properties of Human Driving Behavior Using Multi-Agent Reward Augmented Imitation Learning, ICRA 2019, [paper] [code]
Learning from Demonstration in the Wild, ICRA 2018, [paper]
Multi-Agent Connected Autonomous Driving using Deep Reinforcement Learning, NeurIPS 2019, [paper] [code]
Model-free Deep Reinforcement Learning for Urban Autonomous Driving, ITSC 2019, [paper]
A reinforcement learning based approach for automated lane change maneuvers, IV 2018, [paper]
Adversarial Inverse Reinforcement Learning for Decision Making in Autonomous Driving, ICRA 2020, [paper]
Deep hierarchical reinforcement learning for autonomous driving with distinct behaviors, IV 2018, [paper]
A Hierarchical Architecture for Sequential Decision-Making in Autonomous Driving using Deep Reinforcement Learning, ICML 2019, [paper]
End-to-end Interpretable Neural Motion Planner, CVPR 2019, [paper]
Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles, IROS 2019, [paper]
Dynamic Input for Deep Reinforcement Learning in Autonomous Driving, IROS 2019, [paper]
Learning to Navigate in Cities Without a Map, NIPS 2018, [paper]
Scalable End-to-End Autonomous Vehicle Testing via Rare-event Simulation, NIPS 2018, [paper]
Data-Efficient Hierarchical Reinforcement Learning, NIPS 2018, [paper]
vae+PPO in carla, Accelerating Training of Deep Reinforcement Learning-based Autonomous Driving Agents Through Comparative Study of Agent and Environment Designs, 2019, paper,code
Autonomous driving at intersections: a critical-turning-point approach for left turn, 2020, paper
TPNet: Trajectory Proposal Network for Motion Prediction, 2020, CVPR, paper
DRL+PGM, Interpretable End-to-end Urban Autonomous Driving with Latent Deep Reinforcement Learning, 2020, paper
#Urban Autonomous Driving
Dreaming about Driving, 2018, Wayve, blog
Recurrent World Models Facilitate Policy Evolution, 2018, NIPS, paper, code, blog
Model-Predictive Policy Learning with Uncertainty Regularization for Driving in Dense Traffic， ICLR, 2019, paper, code, video
Variational End-to-End Navigation and Localization, ICRA, 2019, paper
Urban Driving with Conditional Imitation Learning, Wayve, 2019, paper
Learning to Drive from Simulation without Real World Labels, Wayve, 2018, paper
End-to-End Model-Free Reinforcement Learning for Urban Driving using Implicit Affordances, valeo, 2019, paper
OUR TOP TIPS FOR CONDUCTING ROBOTICS FIELD RESEARCH, 2019, blog
Urban Driving with Multi-Objective Deep Reinforcement Learning, AMMAS, paper
DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars, 2018, paper
DeepXplore: Automated Whitebox Testing of Deep Learning Systems, 2017, paper
Driving Policy Transfer via Modularity and Abstraction, 2018, CoRL paper
-
combined prediction with planning under uncertainty / Model-free & model-based
-
combined NN with rules to guarantee safey
-
desicion model testing to evaluate the performance
1.1. Learning by cheating， 2019， paper, code
该论文直接输出轨迹，动作空间可以选用，输入为带有command的local map
1.2. Learning Driving decisions by imitating drivers' control behaviors, 2019, [paper] (https://arxiv.org/pdf/1912.00191.pdf)
该论文的动作空间选用了带速度规划的goal作为离散动作空间，可以参考，输入为驾驶员视野的图像，利用的是GAIL的思路，解决了控制部分独立情况下的梯度反向问题，重参数化技巧。
1.3. Probabilistic Future Prediction for Video Scene Understanding, wayve, 2020 paper
基于预测下的决策： 场景理解与预测，预测未来的分割，深度，光流信息，数据来源CityScapes[13], Mapillary Vistas [46], ApolloScape [28] and Berkeley Deep Drive [65]. 语义分割的未来预测，在cityscape上测试
1.4. 无保护左转的论文阅读与梳理，从规控的角度看预测（高精度地图下的local map 和 感知范围内的local map可以不同），基于预测做如何做决策
1.6. world model系列，在PPUU，Interpretable E2E， world-model，planet，Dreamer， wayve情境预测 的基础上理解world model的运作机理，为planning中的快慢线中的慢线提供支撑
2.1. DRL+PGM, Interpretable End-to-end Urban Autonomous Driving with Latent Deep Reinforcement Learning, 2020, paper, code
该论文的状态空间可以直接选用， 将PGM和world model与 RL相结合，结合了PPUU的隐空间架构，属于Model-based 与 model-free的结合
2.2. DeepXplore: Automated Whitebox Testing of Deep Learning Systems, 2017, paper， code, deeptest code
该论文为可解释性中的loss设计提供了好的思路，考虑来结合对抗测试的方式来提升算法的性能，同时为planning中的快慢线中的快线提供支撑
2.3. Self-Supervised Discovering of Causal Features: Towards Interpretable Reinforcement Learning, 清华大学， 2020, paper
our framework learns to predict an attention mask to highlight the features that may be task-relevant in the state. 为了引入可解释性，如何选择可解释性的loss参考
2.4. Towards Interpretable Reinforcement Learning Using Attention Augmented Agents， DeepMind，Nips，2019, paper
简介： soft-attention model for RL： 关注任务相关的信息， attention的输出是 可视化信息， 关注space和context两个方面,论文工作很好，但是目前缺少结合的思路，值得后续继续再深入理解
how decisions are taken, what information is used and why mistakes are made. tracking the region ahead of the player, focusing on enemies and important moving objects
2.5 Toward Driving Scene Understanding: A Dataset for Learning Driver Behavior and Causal Reasoning, CVPR, 2018 paper
简介：提供了140小时驾驶员数据集，we need to understand the interactions between human driver behaviors and
the corresponding traffic scene situations，视频数据+位置车速等信息，定义了4类标签：goal-directed驾驶意图(左转/右转等)，应激性行为(停车/偏离车道)，应激性行为的原因cause，attention目标
2.6. graph attention 机制的融合，可解释性论文的持续阅读，特别是基于数据的IL下的可解释性
2.7 Transparency and Explanation in Deep Reinforcement Learning Neural Networks, 2018, AIES, paper
2.8 Generation of Policy-Level Explanations for Reinforcement Learning, 2018, AAAI, paper
2.9 Explaining Decisions of a Deep Reinforcement Learner with a Cognitive Architecture, 2018, paper
Social Attention for Autonomous Decision-Making in Dense Traffic， 2019， paper code blog
-
gym-carla环境的安装与掌握 code 解决环境配置的问题
-
local-map生成的掌握 code 解决输入状态表征的问题
-
baseline的掌握 [PPO/SAC code] 解决实验对比的问题
-
Learning by cheating code解决动作空间选择的问题
-
Interpretable E2E code 解决world-model的问题
-
graph attention的引入DRL
-
prediction-dependent decision的引入
-
world model 的引入
-
deep test的引入
End-to-end Interpretable Neural Motion Planner, CVPR 2019, [paper]
Motion Prediction of Traffic Actors for Autonomous Driving using Deep Convolutional Networks， 2018， paper
Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles, IROS 2019, [paper]
Dynamic Input for Deep Reinforcement Learning in Autonomous Driving, IROS 2019, [paper]
Learning to Navigate in Cities Without a Map, NIPS 2018, [paper]
Scalable End-to-End Autonomous Vehicle Testing via Rare-event Simulation, NIPS 2018, [paper]
Deep Imitative Models for Flexible Inference, Planning, and Control, ICLR, 2020 paper, code
MultiPath: Multiple Probabilistic Anchor Trajectory Hypotheses for Behavior Prediction, 2019, CORL paper
AGIL: Learning Attention from Human for Visuomotor Tasks， 2018， paper
Skill Transfer in Deep Reinforcement Learning under Morphological Heterogeneity, TNNLS, 2020
Model-free Deep Reinforcement Learning for Urban Autonomous Driving, ITSC 2019, [paper]
End-to-End Model-Free Reinforcement Learning for Urban Driving using Implicit Affordances, valeo, 2019, paper
Learning Robust Control Policies for End-to-End Autonomous Driving From Data-Driven Simulation, 2020, paper
待阅读， 可解释性与skill transfer以及潜空间之间有着非常紧密的联系