基于变压器神经网络体系结构的自然语言处理（NLP）的令人印象深刻的结果激发了研究人员探索视线离线增强学习（RL）作为通用序列建模问题。基于此范式的最新著作已获得最新的结果，其中一些主要确定性的离线Atari和D4RL基准。但是，由于这些方法将国家和行动共同模拟单一的测序问题，因此它们努力将政策和世界动态对回报的影响解散。因此，在对抗或随机环境中，这些方法导致过度乐观的行为，在自主驾驶（例如自主驾驶）中可能是危险的。在这项工作中，我们提出了一种通过明确解开政策和世界模型来解决这种乐观偏见的方法，该方法使我们在测试时可以搜索对环境中多个可能的未来的稳健性的策略。我们在模拟中的各种自动驾驶任务上展示了我们的方法的出色性能。

尽管基于计划的序列建模方法在连续控制方面表现出巨大的潜力，但由于高维空间中规划的高度计算复杂性和天生的困难，将它们扩展到高维状态序列仍然是一个开放的挑战。我们提出了轨迹自动编码计划器（TAP），这是一种基于计划的序列建模RL方法，可扩展到高州行动维度。使用状态条件矢量定量的变分自动编码器（VQ-VAE），点击模拟给定当前状态的轨迹的条件分布。当部署为RL代理时，TAP避免在高维连续动作空间中逐步计划，而是通过Beam Search寻找最佳的潜在代码序列。与$ o（d^3）$轨迹变压器的复杂性不同，TAP享受常数$ o（c）$规划有关州行动维度$ d $的计算复杂性。我们的经验评估还表明，随着维度的增长，TAP的表现越来越强。对于具有较高状态和动作维度的ADROIT机器人手动操纵任务，TAP超过了基于模型的方法，包括TT，其边距很大，并且还击败了强大的无模型参与者 - 批评基准。

Transformer, originally devised for natural language processing, has also attested significant success in computer vision. Thanks to its super expressive power, researchers are investigating ways to deploy transformers to reinforcement learning (RL) and the transformer-based models have manifested their potential in representative RL benchmarks. In this paper, we collect and dissect recent advances on transforming RL by transformer (transformer-based RL or TRL), in order to explore its development trajectory and future trend. We group existing developments in two categories: architecture enhancement and trajectory optimization, and examine the main applications of TRL in robotic manipulation, text-based games, navigation and autonomous driving. For architecture enhancement, these methods consider how to apply the powerful transformer structure to RL problems under the traditional RL framework, which model agents and environments much more precisely than deep RL methods, but they are still limited by the inherent defects of traditional RL algorithms, such as bootstrapping and "deadly triad". For trajectory optimization, these methods treat RL problems as sequence modeling and train a joint state-action model over entire trajectories under the behavior cloning framework, which are able to extract policies from static datasets and fully use the long-sequence modeling capability of the transformer. Given these advancements, extensions and challenges in TRL are reviewed and proposals about future direction are discussed. We hope that this survey can provide a detailed introduction to TRL and motivate future research in this rapidly developing field.

离线强化学习（RL）为从离线数据提供学习决策的框架，因此构成了现实世界应用程序作为自动驾驶的有希望的方法。自动驾驶车辆（SDV）学习策略，这甚至可能甚至优于次优数据集中的行为。特别是在安全关键应用中，作为自动化驾驶，解释性和可转换性是成功的关键。这激发了使用基于模型的离线RL方法，该方法利用规划。然而，目前的最先进的方法往往忽视了多种子体系统随机行为引起的溶液不确定性的影响。这项工作提出了一种新的基于不确定感知模型的离线强化学习利用规划（伞）的新方法，其解决了以可解释的基于学习的方式共同的预测，规划和控制问题。训练有素的动作调节的随机动力学模型捕获了交通场景的独特不同的未来演化。分析为我们在挑战自动化驾驶模拟中的效力和基于现实世界的公共数据集的方法提供了经验证据。

With the development of deep representation learning, the domain of reinforcement learning (RL) has become a powerful learning framework now capable of learning complex policies in high dimensional environments. This review summarises deep reinforcement learning (DRL) algorithms and provides a taxonomy of automated driving tasks where (D)RL methods have been employed, while addressing key computational challenges in real world deployment of autonomous driving agents. It also delineates adjacent domains such as behavior cloning, imitation learning, inverse reinforcement learning that are related but are not classical RL algorithms. The role of simulators in training agents, methods to validate, test and robustify existing solutions in RL are discussed.

Behavioural cloning (BC) is a commonly used imitation learning method to infer a sequential decision-making policy from expert demonstrations. However, when the quality of the data is not optimal, the resulting behavioural policy also performs sub-optimally once deployed. Recently, there has been a surge in offline reinforcement learning methods that hold the promise to extract high-quality policies from sub-optimal historical data. A common approach is to perform regularisation during training, encouraging updates during policy evaluation and/or policy improvement to stay close to the underlying data. In this work, we investigate whether an offline approach to improving the quality of the existing data can lead to improved behavioural policies without any changes in the BC algorithm. The proposed data improvement approach - Trajectory Stitching (TS) - generates new trajectories (sequences of states and actions) by `stitching' pairs of states that were disconnected in the original data and generating their connecting new action. By construction, these new transitions are guaranteed to be highly plausible according to probabilistic models of the environment, and to improve a state-value function. We demonstrate that the iterative process of replacing old trajectories with new ones incrementally improves the underlying behavioural policy. Extensive experimental results show that significant performance gains can be achieved using TS over BC policies extracted from the original data. Furthermore, using the D4RL benchmarking suite, we demonstrate that state-of-the-art results are obtained by combining TS with two existing offline learning methodologies reliant on BC, model-based offline planning (MBOP) and policy constraint (TD3+BC).

由于互动交通参与者的随机性质和道路结构的复杂性，城市自动驾驶的决策是具有挑战性的。尽管基于强化的学习（RL）决策计划有望处理城市驾驶方案，但它的样本效率低和适应性差。在本文中，我们提出了Scene-Rep Transformer，以通过更好的场景表示编码和顺序预测潜在蒸馏来提高RL决策能力。具体而言，构建了多阶段变压器（MST）编码器，不仅对自我车辆及其邻居之间的相互作用意识进行建模，而且对代理商及其候选路线之间的意图意识。具有自我监督学习目标的连续潜伏变压器（SLT）用于将未来的预测信息提炼成潜在的场景表示，以减少勘探空间并加快训练的速度。基于软演员批评的最终决策模块（SAC）将来自场景rep变压器的精制潜在场景表示输入，并输出驾驶动作。该框架在五个挑战性的模拟城市场景中得到了验证，其性能通过成功率，安全性和效率方面的数据效率和性能的大幅度提高来定量表现出来。定性结果表明，我们的框架能够提取邻居代理人的意图，以帮助做出决策并提供更多多元化的驾驶行为。

强化学习（RL）通常涉及估计静止政策或单步模型，利用马尔可夫属性来解决问题。但是，我们也可以将RL视为通用序列建模问题，目标是产生一系列导致一系列高奖励的动作。通过这种方式观看，考虑在其他域中运用良好的高容量序列预测模型，例如自然语言处理，也可以为RL问题提供有效的解决方案。为此，我们探索如何使用变压器架构与序列建模的工具来解决RL，以将分布在轨迹上和将光束搜索作为规划算法进行重新定位。框架RL作为序列建模问题简化了一系列设计决策，允许我们分配在离线RL算法中常见的许多组件。我们展示了这种方法跨越长地平动态预测，仿制学习，目标条件的RL和离线RL的灵活性。此外，我们表明这种方法可以与现有的无模型算法结合起来，以在稀疏奖励，长地平线任务中产生最先进的策划仪。

Designing a safe and human-like decision-making system for an autonomous vehicle is a challenging task. Generative imitation learning is one possible approach for automating policy-building by leveraging both real-world and simulated decisions. Previous work that applies generative imitation learning to autonomous driving policies focuses on learning a low-level controller for simple settings. However, to scale to complex settings, many autonomous driving systems combine fixed, safe, optimization-based low-level controllers with high-level decision-making logic that selects the appropriate task and associated controller. In this paper, we attempt to bridge this gap in complexity by employing Safety-Aware Hierarchical Adversarial Imitation Learning (SHAIL), a method for learning a high-level policy that selects from a set of low-level controller instances in a way that imitates low-level driving data on-policy. We introduce an urban roundabout simulator that controls non-ego vehicles using real data from the Interaction dataset. We then demonstrate empirically that even with simple controller options, our approach can produce better behavior than previous approaches in driver imitation that have difficulty scaling to complex environments. Our implementation is available at https://github.com/sisl/InteractionImitation.

在离线强化学习（离线RL）中，主要挑战之一是处理学习策略与给定数据集之间的分布转变。为了解决这个问题，最近的离线RL方法试图引入保守主义偏见，以鼓励在高信心地区学习。无模型方法使用保守的正常化或特殊网络结构直接对策略或价值函数学习进行这样的偏见，但它们约束的策略搜索限制了脱机数据集之外的泛化。基于模型的方法使用保守量量化学习前瞻性动态模型，然后生成虚构的轨迹以扩展脱机数据集。然而，由于离线数据集中的有限样本，保守率量化通常在支撑区域内遭受全面化。不可靠的保守措施将误导基于模型的想象力，以不受欢迎的地区，导致过多的行为。为了鼓励更多的保守主义，我们提出了一种基于模型的离线RL框架，称为反向离线模型的想象（ROMI）。我们与新颖的反向策略结合使用逆向动力学模型，该模型可以生成导致脱机数据集中的目标目标状态的卷展栏。这些反向的想象力提供了无通知的数据增强，以便无模型策略学习，并使远程数据集的保守概括。 ROMI可以有效地与现成的无模型算法组合，以实现基于模型的概括，具有适当的保守主义。经验结果表明，我们的方法可以在离线RL基准任务中产生更保守的行为并实现最先进的性能。

Recent improvements in conditional generative modeling have made it possible to generate high-quality images from language descriptions alone. We investigate whether these methods can directly address the problem of sequential decision-making. We view decision-making not through the lens of reinforcement learning (RL), but rather through conditional generative modeling. To our surprise, we find that our formulation leads to policies that can outperform existing offline RL approaches across standard benchmarks. By modeling a policy as a return-conditional diffusion model, we illustrate how we may circumvent the need for dynamic programming and subsequently eliminate many of the complexities that come with traditional offline RL. We further demonstrate the advantages of modeling policies as conditional diffusion models by considering two other conditioning variables: constraints and skills. Conditioning on a single constraint or skill during training leads to behaviors at test-time that can satisfy several constraints together or demonstrate a composition of skills. Our results illustrate that conditional generative modeling is a powerful tool for decision-making.

Offline reinforcement-learning (RL) algorithms learn to make decisions using a given, fixed training dataset without the possibility of additional online data collection. This problem setting is captivating because it holds the promise of utilizing previously collected datasets without any costly or risky interaction with the environment. However, this promise also bears the drawback of this setting. The restricted dataset induces subjective uncertainty because the agent can encounter unfamiliar sequences of states and actions that the training data did not cover. Moreover, inherent system stochasticity further increases uncertainty and aggravates the offline RL problem, preventing the agent from learning an optimal policy. To mitigate the destructive uncertainty effects, we need to balance the aspiration to take reward-maximizing actions with the incurred risk due to incorrect ones. In financial economics, modern portfolio theory (MPT) is a method that risk-averse investors can use to construct diversified portfolios that maximize their returns without unacceptable levels of risk. We integrate MPT into the agent's decision-making process to present a simple-yet-highly-effective risk-aware planning algorithm for offline RL. Our algorithm allows us to systematically account for the \emph{estimated quality} of specific actions and their \emph{estimated risk} due to the uncertainty. We show that our approach can be coupled with the Transformer architecture to yield a state-of-the-art planner for offline RL tasks, maximizing the return while significantly reducing the variance.

我们解决了由具有不同驱动程序行为的道路代理人填充的密集模拟交通环境中的自我车辆导航问题。由于其异构行为引起的代理人的不可预测性，这种环境中的导航是挑战。我们提出了一种新的仿真技术，包括丰富现有的交通模拟器，其具有与不同程度的侵略性程度相对应的行为丰富的轨迹。我们在驾驶员行为建模算法的帮助下生成这些轨迹。然后，我们使用丰富的模拟器培训深度加强学习（DRL）策略，包括一组高级车辆控制命令，并在测试时间使用此策略来执行密集流量的本地导航。我们的政策隐含地模拟了交通代理商之间的交互，并计算了自助式驾驶员机动，例如超速，超速，编织和突然道路变化的激进驾驶员演习的安全轨迹。我们增强的行为丰富的模拟器可用于生成由对应于不同驱动程序行为和流量密度的轨迹组成的数据集，我们的行为的导航方案可以与最先进的导航算法相结合。

While large-scale sequence modeling from offline data has led to impressive performance gains in natural language and image generation, directly translating such ideas to robotics has been challenging. One critical reason for this is that uncurated robot demonstration data, i.e. play data, collected from non-expert human demonstrators are often noisy, diverse, and distributionally multi-modal. This makes extracting useful, task-centric behaviors from such data a difficult generative modeling problem. In this work, we present Conditional Behavior Transformers (C-BeT), a method that combines the multi-modal generation ability of Behavior Transformer with future-conditioned goal specification. On a suite of simulated benchmark tasks, we find that C-BeT improves upon prior state-of-the-art work in learning from play data by an average of 45.7%. Further, we demonstrate for the first time that useful task-centric behaviors can be learned on a real-world robot purely from play data without any task labels or reward information. Robot videos are best viewed on our project website: https://play-to-policy.github.io

在本文中，我们提出了一个系统，以培训不仅从自我车辆收集的经验，而且还观察到的所有车辆的经验。该系统使用其他代理的行为来创建更多样化的驾驶场景，而无需收集其他数据。从其他车辆学习的主要困难是没有传感器信息。我们使用一组监督任务来学习一个中间表示，这是对控制车辆的观点不变的。这不仅在训练时间提供了更丰富的信号，而且还可以在推断过程中进行更复杂的推理。了解所有车辆驾驶如何有助于预测测试时的行为，并避免碰撞。我们在闭环驾驶模拟中评估该系统。我们的系统的表现优于公共卡拉排行榜上的所有先前方法，较大的利润率将驾驶得分提高了25，路线完成率提高了24分。我们的方法赢得了2021年的卡拉自动驾驶挑战。代码和数据可在https://github.com/dotchen/lav上获得。

离线强化学习在利用大型预采用的数据集进行政策学习方面表现出了巨大的希望，使代理商可以放弃经常廉价的在线数据收集。但是，迄今为止，离线强化学习的探索相对较小，并且缺乏对剩余挑战所在的何处的了解。在本文中，我们试图建立简单的基线以在视觉域中连续控制。我们表明，对两个基于最先进的在线增强学习算法，Dreamerv2和DRQ-V2进行了简单的修改，足以超越事先工作并建立竞争性的基准。我们在现有的离线数据集中对这些算法进行了严格的评估，以及从视觉观察结果中进行离线强化学习的新测试台，更好地代表现实世界中离线增强学习问题中存在的数据分布，并开放我们的代码和数据以促进此方面的进度重要领域。最后，我们介绍并分析了来自视觉观察的离线RL所独有的几个关键Desiderata，包括视觉分散注意力和动态视觉上可识别的变化。

ML-based motion planning is a promising approach to produce agents that exhibit complex behaviors, and automatically adapt to novel environments. In the context of autonomous driving, it is common to treat all available training data equally. However, this approach produces agents that do not perform robustly in safety-critical settings, an issue that cannot be addressed by simply adding more data to the training set - we show that an agent trained using only a 10% subset of the data performs just as well as an agent trained on the entire dataset. We present a method to predict the inherent difficulty of a driving situation given data collected from a fleet of autonomous vehicles deployed on public roads. We then demonstrate that this difficulty score can be used in a zero-shot transfer to generate curricula for an imitation-learning based planning agent. Compared to training on the entire unbiased training dataset, we show that prioritizing difficult driving scenarios both reduces collisions by 15% and increases route adherence by 14% in closed-loop evaluation, all while using only 10% of the training data.

Making safe and human-like decisions is an essential capability of autonomous driving systems and learning-based behavior planning is a promising pathway toward this objective. Distinguished from existing learning-based methods that directly output decisions, this work introduces a predictive behavior planning framework that learns to predict and evaluate from human driving data. Concretely, a behavior generation module first produces a diverse set of candidate behaviors in the form of trajectory proposals. Then the proposed conditional motion prediction network is employed to forecast other agents' future trajectories conditioned on each trajectory proposal. Given the candidate plans and associated prediction results, we learn a scoring module to evaluate the plans using maximum entropy inverse reinforcement learning (IRL). We conduct comprehensive experiments to validate the proposed framework on a large-scale real-world urban driving dataset. The results reveal that the conditional prediction model is able to forecast multiple possible future trajectories given a candidate behavior and the prediction results are reactive to different plans. Moreover, the IRL-based scoring module can properly evaluate the trajectory proposals and select close-to-human ones. The proposed framework outperforms other baseline methods in terms of similarity to human driving trajectories. Moreover, we find that the conditional prediction model can improve both prediction and planning performance compared to the non-conditional model, and learning the scoring module is critical to correctly evaluating the candidate plans to align with human drivers.

人类行为的不确定性对拥挤的城市环境中的自动驾驶构成了重大挑战。部分可观察到的马尔可夫决策过程（POMDP）为不确定性下的计划提供了一个原则的框架，通常利用蒙特卡洛抽样来实现在线绩效进行复杂的任务。但是，抽样还通过潜在缺失关键事件引起了安全问题。为了解决这个问题，我们提出了一种新的算法，学习对驾驶行为（领导者）的关注，这些算法在计划过程中学习了批判性人类行为。领导者学习了一个神经网络生成器，以实时情况下对人类行为的关注。它将注意力集成到信仰空间计划者中，使用重要性抽样来偏向关键事件。为了训练该算法，我们让注意力生成器和计划者组成了最小游戏。通过解决Min-Max游戏，领导者学会了无需人类标签即可执行风险意识的计划。

End-to-end autonomous driving provides a feasible way to automatically maximize overall driving system performance by directly mapping the raw pixels from a front-facing camera to control signals. Recent advanced methods construct a latent world model to map the high dimensional observations into compact latent space. However, the latent states embedded by the world model proposed in previous works may contain a large amount of task-irrelevant information, resulting in low sampling efficiency and poor robustness to input perturbations. Meanwhile, the training data distribution is usually unbalanced, and the learned policy is hard to cope with the corner cases during the driving process. To solve the above challenges, we present a semantic masked recurrent world model (SEM2), which introduces a latent filter to extract key task-relevant features and reconstruct a semantic mask via the filtered features, and is trained with a multi-source data sampler, which aggregates common data and multiple corner case data in a single batch, to balance the data distribution. Extensive experiments on CARLA show that our method outperforms the state-of-the-art approaches in terms of sample efficiency and robustness to input permutations.