An efficient and reliable multi-agent decision-making system is highly demanded for the safe and efficient operation of connected autonomous vehicles in intelligent transportation systems. Current researches mainly focus on the Deep Reinforcement Learning (DRL) methods. However, utilizing DRL methods in interactive traffic scenarios is hard to represent the mutual effects between different vehicles and model the dynamic traffic environments due to the lack of interactive information in the representation of the environments, which results in low accuracy of cooperative decisions generation. To tackle these difficulties, this research proposes a framework to enable different Graph Reinforcement Learning (GRL) methods for decision-making, and compares their performance in interactive driving scenarios. GRL methods combinate the Graph Neural Network (GNN) and DRL to achieve the better decisions generation in interactive scenarios of autonomous vehicles, where the features of interactive scenarios are extracted by the GNN, and cooperative behaviors are generated by DRL framework. Several GRL approaches are summarized and implemented in the proposed framework. To evaluate the performance of the proposed GRL methods, an interactive driving scenarios on highway with two ramps is constructed, and simulated experiment in the SUMO platform is carried out to evaluate the performance of different GRL approaches. Finally, results are analyzed in multiple perspectives and dimensions to compare the characteristic of different GRL approaches in intelligent transportation scenarios. Results show that the implementation of GNN can well represents the interaction between vehicles, and the combination of GNN and DRL is able to improve the performance of the generation of lane-change behaviors. The source code of our work can be found at https://github.com/Jacklinkk/TorchGRL.

We introduce ApolloRL, an open platform for research in reinforcement learning for autonomous driving. The platform provides a complete closed-loop pipeline with training, simulation, and evaluation components. It comes with 300 hours of real-world data in driving scenarios and popular baselines such as Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) agents. We elaborate in this paper on the architecture and the environment defined in the platform. In addition, we discuss the performance of the baseline agents in the ApolloRL environment.

Unfortunately, conventional approaches pay little attention to most defects [Fung In federated learning (FL), model aggregation has et al., 2018]. Therefore, an efficient approach to alleviating been widely adopted for data privacy. In recent performance degradation caused by defective local models is years, assigning different weights to local models strongly needed for FL. Existing researches on blockchainbased has been used to alleviate the FL performance FL have defined the concept of reputation, which manifests degradation caused by differences between local the reliability of each local model [Kang et al., 2019] datasets. However, when various defects make the [Kang et al., 2020]. Similarly, we evaluate the model quality FL process unreliable, most existing FL approaches to measure how trustworthy a local model is. After learning expose weak robustness. In this paper, we propose about the quality of each local model, we are motivated to design the DEfect-AwaRe federated soft actor-critic a deep neural network (DNN) to assign optimal weights (DearFSAC) to dynamically assign weights to local to local models, so that the global model can maintain a considerable models to improve the robustness of FL. The deep performance no matter if there exist defects or not.

Although machine learning (ML) has been successful in automating various software engineering needs, software testing still remains a highly challenging topic. In this paper, we aim to improve the generative testing of software by directly augmenting the random number generator (RNG) with a deep reinforcement learning (RL) agent using an efficient, automatically extractable state representation of the software under test. Using the Cosmos SDK as the testbed, we show that the proposed DeepRNG framework provides a statistically significant improvement to the testing of the highly complex software library with over 350,000 lines of code. The source code of the DeepRNG framework is publicly available online.

The last several years have witnessed the great success of reinforcement learning (RL) including the video game playing [Mnih et al., 2015], robot manipulation [Gu et al., 2017], autonomous driving [Shalev-Shwartz et al., 2016] and many others [Lazic et al., 2018, Dalal et al., 2016]. Most of them focus on the problem where the system of interest evolves continuously with time, e.g., a trajectory of a tennis ball. However, the conventional research in RL may omit a category of system that evolves continuously and may be interrupted by stochastic events abruptly (see the jumps in Figure 1). Such system exists ubiquitously in the social and information science and therefore necessitates the research of reinforcement learning in these domains to extend its applicability in the real-world problems [Farajtabar et al., 2017, Wang et al., 2018], in which the agent seeks an optimal intervention policy so as to improve the future course of events. Concrete examples may include: - Social media. Social media website allows users to create and share content. Retweet can form as users resharing and broadcasting others' tweet to their friends and followers. Such stochastic events would steer the behaviors of other tweet users [Rizoiu et al., 2017]. At the same time, the platform (agent) may want to seek a policy to effectively mitigate the fake news by optimizing the performance of real news propagation over the network Farajtabar et al. [2017].

Zeroth-order optimization methods and policy gradient based first-order methods are two promising alternatives to solve reinforcement learning (RL) problems with complementary advantages. The former work with arbitrary policies, drive state-dependent and temporally-extended exploration, possess robustness-seeking property, but suffer from high sample complexity, while the latter are more sample efficient but restricted to differentiable policies and the learned policies are less robust. We propose Zeroth-Order Actor-Critic algorithm (ZOAC) that unifies these two methods into an on-policy actor-critic architecture to preserve the advantages from both. ZOAC conducts rollouts collection with timestep-wise perturbation in parameter space, first-order policy evaluation (PEV) and zeroth-order policy improvement (PIM) alternately in each iteration. We evaluate our proposed method on a range of challenging continuous control benchmarks using different types of policies, where ZOAC outperforms zeroth-order and first-order baseline algorithms.

We consider offline Imitation Learning from corrupted demonstrations where a constant fraction of data can be noise or even arbitrary outliers. Classical approaches such as Behavior Cloning assumes that demonstrations are collected by an presumably optimal expert, hence may fail drastically when learning from corrupted demonstrations. We propose a novel robust algorithm by minimizing a Median-of-Means (MOM) objective which guarantees the accurate estimation of policy, even in the presence of constant fraction of outliers. Our theoretical analysis shows that our robust method in the corrupted setting enjoys nearly the same error scaling and sample complexity guarantees as the classical Behavior Cloning in the expert demonstration setting. Our experiments on continuous-control benchmarks validate that our method exhibits the predicted robustness and effectiveness, and achieves competitive results compared to existing imitation learning methods.

Deep reinforcement learning (DRL) has attracted much attention as an approach to solve sequential decision making problems without mathematical models of systems or environments. In general, a constraint may be imposed on the decision making. In this study, we consider the optimal decision making problems with constraints to complete temporal high-level tasks in the continuous state-action domain. We describe the constraints using signal temporal logic (STL), which is useful for time sensitive control tasks since it can specify continuous signals within a bounded time interval. To deal with the STL constraints, we introduce an extended constrained Markov decision process (CMDP), which is called a $\tau$-CMDP. We formulate the STL constrained optimal decision making problem as the $\tau$-CMDP and propose a two-phase constrained DRL algorithm using the Lagrangian relaxation method. Through simulations, we also demonstrate the learning performance of the proposed algorithm.

In order for humans to confidently decide where to employ RL agents for real-world tasks, a human developer must validate that the agent will perform well at test-time. Some policy interpretability methods facilitate this by capturing the policy's decision making in a set of agent rollouts. However, even the most informative trajectories of training time behavior may give little insight into the agent's behavior out of distribution. In contrast, our method conveys how the agent performs under distribution shifts by showing the agent's behavior across a wider trajectory distribution. We generate these trajectories by guiding the agent to more diverse unseen states and showing the agent's behavior there. In a user study, we demonstrate that our method enables users to score better than baseline methods on one of two agent validation tasks.

Cooperative artificial intelligence with human or superhuman proficiency in collaborative tasks stands at the frontier of machine learning research. Prior work has tended to evaluate cooperative AI performance under the restrictive paradigms of self-play (teams composed of agents trained together) and cross-play (teams of agents trained independently but using the same algorithm). Recent work has indicated that AI optimized for these narrow settings may make for undesirable collaborators in the real-world. We formalize an alternative criteria for evaluating cooperative AI, referred to as inter-algorithm cross-play, where agents are evaluated on teaming performance with all other agents within an experiment pool with no assumption of algorithmic similarities between agents. We show that existing state-of-the-art cooperative AI algorithms, such as Other-Play and Off-Belief Learning, under-perform in this paradigm. We propose the Any-Play learning augmentation -- a multi-agent extension of diversity-based intrinsic rewards for zero-shot coordination (ZSC) -- for generalizing self-play-based algorithms to the inter-algorithm cross-play setting. We apply the Any-Play learning augmentation to the Simplified Action Decoder (SAD) and demonstrate state-of-the-art performance in the collaborative card game Hanabi.