Jiacheng Zhu

Haohong Lin, Wenhao Ding, Zuxin Liu et al. · cmu

In the domain of autonomous driving, the offline Reinforcement Learning~(RL) approaches exhibit notable efficacy in addressing sequential decision-making problems from offline datasets. However, maintaining safety in diverse safety-critical scenarios remains a significant challenge due to long-tailed and unforeseen scenarios absent from offline datasets. In this paper, we introduce the saFety-aware strUctured Scenario representatION (FUSION), a pioneering representation learning method in offline RL to facilitate the learning of a generalizable end-to-end driving policy by leveraging structured scenario information. FUSION capitalizes on the causal relationships between the decomposed reward, cost, state, and action space, constructing a framework for structured sequential reasoning in dynamic traffic environments. We conduct extensive evaluations in two typical real-world settings of the distribution shift in autonomous vehicles, demonstrating the good balance between safety cost and utility reward compared to the current state-of-the-art safe RL and IL baselines. Empirical evidence in various driving scenarios attests that FUSION significantly enhances the safety and generalizability of autonomous driving agents, even in the face of challenging and unseen environments. Furthermore, our ablation studies reveal noticeable improvements in the integration of causal representation into the offline safe RL algorithm. Our code implementation is available at: https://sites.google.com/view/safe-fusion/.

Ajay Patel, Jiacheng Zhu, Justin Qiu et al.

Style representations aim to embed texts with similar writing styles closely and texts with different styles far apart, regardless of content. However, the contrastive triplets often used for training these representations may vary in both style and content, leading to potential content leakage in the representations. We introduce StyleDistance, a novel approach to training stronger content-independent style embeddings. We use a large language model to create a synthetic dataset of near-exact paraphrases with controlled style variations, and produce positive and negative examples across 40 distinct style features for precise contrastive learning. We assess the quality of our synthetic data and embeddings through human and automatic evaluations. StyleDistance enhances the content-independence of style embeddings, which generalize to real-world benchmarks and outperform leading style representations in downstream applications. Our model can be found at https://huggingface.co/StyleDistance/styledistance .

Bertie Vidgen, Adarsh Agrawal, Ahmed M. Ahmed et al. · deepmind, oxford

This paper introduces v0.5 of the AI Safety Benchmark, which has been created by the MLCommons AI Safety Working Group. The AI Safety Benchmark has been designed to assess the safety risks of AI systems that use chat-tuned language models. We introduce a principled approach to specifying and constructing the benchmark, which for v0.5 covers only a single use case (an adult chatting to a general-purpose assistant in English), and a limited set of personas (i.e., typical users, malicious users, and vulnerable users). We created a new taxonomy of 13 hazard categories, of which 7 have tests in the v0.5 benchmark. We plan to release version 1.0 of the AI Safety Benchmark by the end of 2024. The v1.0 benchmark will provide meaningful insights into the safety of AI systems. However, the v0.5 benchmark should not be used to assess the safety of AI systems. We have sought to fully document the limitations, flaws, and challenges of v0.5. This release of v0.5 of the AI Safety Benchmark includes (1) a principled approach to specifying and constructing the benchmark, which comprises use cases, types of systems under test (SUTs), language and context, personas, tests, and test items; (2) a taxonomy of 13 hazard categories with definitions and subcategories; (3) tests for seven of the hazard categories, each comprising a unique set of test items, i.e., prompts. There are 43,090 test items in total, which we created with templates; (4) a grading system for AI systems against the benchmark; (5) an openly available platform, and downloadable tool, called ModelBench that can be used to evaluate the safety of AI systems on the benchmark; (6) an example evaluation report which benchmarks the performance of over a dozen openly available chat-tuned language models; (7) a test specification for the benchmark.

Zhongzhi Li, Jingqi Tu, Jiacheng Zhu et al.

Deep transfer learning (DTL) is a fundamental method in the field of Intelligent Fault Detection (IFD). It aims to mitigate the degradation of method performance that arises from the discrepancies in data distribution between training set (source domain) and testing set (target domain). Considering the fact that fault data collection is challenging and certain faults are scarce, DTL-based methods face the limitation of available observable data, which reduces the detection performance of the methods in the target domain. Furthermore, DTL-based methods lack comprehensive uncertainty analysis that is essential for building reliable IFD systems. To address the aforementioned problems, this paper proposes a novel DTL-based method known as Neural Processes-based deep transfer learning with graph convolution network (GTNP). Feature-based transfer strategy of GTNP bridges the data distribution discrepancies of source domain and target domain in high-dimensional space. Both the joint modeling based on global and local latent variables and sparse sampling strategy reduce the demand of observable data in the target domain. The multi-scale uncertainty analysis is obtained by using the distribution characteristics of global and local latent variables. Global analysis of uncertainty enables GTNP to provide quantitative values that reflect the complexity of methods and the difficulty of tasks. Local analysis of uncertainty allows GTNP to model uncertainty (confidence of the fault detection result) at each sample affected by noise and bias. The validation of the proposed method is conducted across 3 IFD tasks, consistently showing the superior detection performance of GTNP compared to the other DTL-based methods.

Mengdi Xu, Peide Huang, Fengpei Li et al.

Evaluating rare but high-stakes events is one of the main challenges in obtaining reliable reinforcement learning policies, especially in large or infinite state/action spaces where limited scalability dictates a prohibitively large number of testing iterations. On the other hand, a biased or inaccurate policy evaluation in a safety-critical system could potentially cause unexpected catastrophic failures during deployment. This paper proposes the Accelerated Policy Evaluation (APE) method, which simultaneously uncovers rare events and estimates the rare event probability in Markov decision processes. The APE method treats the environment nature as an adversarial agent and learns towards, through adaptive importance sampling, the zero-variance sampling distribution for the policy evaluation. Moreover, APE is scalable to large discrete or continuous spaces by incorporating function approximators. We investigate the convergence property of APE in the tabular setting. Our empirical studies show that APE can estimate the rare event probability with a smaller bias while only using orders of magnitude fewer samples than baselines in multi-agent and single-agent environments.

Baiming Chen, Zuxin Liu, Jiacheng Zhu et al.

Safety is a critical concern when deploying reinforcement learning agents for realistic tasks. Recently, safe reinforcement learning algorithms have been developed to optimize the agent's performance while avoiding violations of safety constraints. However, few studies have addressed the non-stationary disturbances in the environments, which may cause catastrophic outcomes. In this paper, we propose the context-aware safe reinforcement learning (CASRL) method, a meta-learning framework to realize safe adaptation in non-stationary environments. We use a probabilistic latent variable model to achieve fast inference of the posterior environment transition distribution given the context data. Safety constraints are then evaluated with uncertainty-aware trajectory sampling. The high cost of safety violations leads to the rareness of unsafe records in the dataset. We address this issue by enabling prioritized sampling during model training and formulating prior safety constraints with domain knowledge during constrained planning. The algorithm is evaluated in realistic safety-critical environments with non-stationary disturbances. Results show that the proposed algorithm significantly outperforms existing baselines in terms of safety and robustness.

Mengdi Xu, Wenhao Ding, Jiacheng Zhu et al.

Continuously learning to solve unseen tasks with limited experience has been extensively pursued in meta-learning and continual learning, but with restricted assumptions such as accessible task distributions, independently and identically distributed tasks, and clear task delineations. However, real-world physical tasks frequently violate these assumptions, resulting in performance degradation. This paper proposes a continual online model-based reinforcement learning approach that does not require pre-training to solve task-agnostic problems with unknown task boundaries. We maintain a mixture of experts to handle nonstationarity, and represent each different type of dynamics with a Gaussian Process to efficiently leverage collected data and expressively model uncertainty. We propose a transition prior to account for the temporal dependencies in streaming data and update the mixture online via sequential variational inference. Our approach reliably handles the task distribution shift by generating new models for never-before-seen dynamics and reusing old models for previously seen dynamics. In experiments, our approach outperforms alternative methods in non-stationary tasks, including classic control with changing dynamics and decision making in different driving scenarios.

Aritra Guha, Rayleigh Lei, Jiacheng Zhu et al.

Robust representation learning of temporal dynamic interactions is an important problem in robotic learning in general and automated unsupervised learning in particular. Temporal dynamic interactions can be described by (multiple) geometric trajectories in a suitable space over which unsupervised learning techniques may be applied to extract useful features from raw and high-dimensional data measurements. Taking a geometric approach to robust representation learning for temporal dynamic interactions, it is necessary to develop suitable metrics and a systematic methodology for comparison and for assessing the stability of an unsupervised learning method with respect to its tuning parameters. Such metrics must account for the (geometric) constraints in the physical world as well as the uncertainty associated with the learned patterns. In this paper we introduce a model-free metric based on the Procrustes distance for robust representation learning of interactions, and an optimal transport based distance metric for comparing between distributions of interaction primitives. These distance metrics can serve as an objective for assessing the stability of an interaction learning algorithm. They are also used for comparing the outcomes produced by different algorithms. Moreover, they may also be adopted as an objective function to obtain clusters and representative interaction primitives. These concepts and techniques will be introduced, along with mathematical properties, while their usefulness will be demonstrated in unsupervised learning of vehicle-to-vechicle interactions extracted from the Safety Pilot database, the world's largest database for connected vehicles.

Chengyuan Zhang, Jiacheng Zhu, Wenshuo Wang et al.

Interpretation of common-yet-challenging interaction scenarios can benefit well-founded decisions for autonomous vehicles. Previous research achieved this using their prior knowledge of specific scenarios with predefined models, limiting their adaptive capabilities. This paper describes a Bayesian nonparametric approach that leverages continuous (i.e., Gaussian processes) and discrete (i.e., Dirichlet processes) stochastic processes to reveal underlying interaction patterns of the ego vehicle with other nearby vehicles. Our model relaxes dependency on the number of surrounding vehicles by developing an acceleration-sensitive velocity field based on Gaussian processes. The experiment results demonstrate that the velocity field can represent the spatial interactions between the ego vehicle and its surroundings. Then, a discrete Bayesian nonparametric model, integrating Dirichlet processes and hidden Markov models, is developed to learn the interaction patterns over the temporal space by segmenting and clustering the sequential interaction data into interpretable granular patterns automatically. We then evaluate our approach in the highway lane-change scenarios using the highD dataset collected from real-world settings. Results demonstrate that our proposed Bayesian nonparametric approach provides an insight into the complicated lane-change interactions of the ego vehicle with multiple surrounding traffic participants based on the interpretable interaction patterns and their transition properties in temporal relationships. Our proposed approach sheds light on efficiently analyzing other kinds of multi-agent interactions, such as vehicle-pedestrian interactions. View the demos via https://youtu.be/z_vf9UHtdAM.

Jiacheng Zhu, Shenghao Qin, Wenshuo Wang et al.

Predicting surrounding vehicle behaviors are critical to autonomous vehicles when negotiating in multi-vehicle interaction scenarios. Most existing approaches require tedious training process with large amounts of data and may fail to capture the propagating uncertainty in interaction behaviors. The multi-vehicle behaviors are assumed to be generated from a stochastic process. This paper proposes an attentive recurrent neural process (ARNP) approach to overcome the above limitations, which uses a neural process (NP) to learn a distribution of multi-vehicle interaction behavior. Our proposed model inherits the flexibility of neural networks while maintaining Bayesian probabilistic characteristics. Constructed by incorporating NPs with recurrent neural networks (RNNs), the ARNP model predicts the distribution of a target vehicle trajectory conditioned on the observed long-term sequential data of all surrounding vehicles. This approach is verified by learning and predicting lane-changing trajectories in complex traffic scenarios. Experimental results demonstrate that our proposed method outperforms previous counterparts in terms of accuracy and uncertainty expressiveness. Moreover, the meta-learning instinct of NPs enables our proposed ARNP model to capture global information of all observations, thereby being able to adapt to new targets efficiently.

Shenghao Qin, Jiacheng Zhu, Jimmy Qin et al.

Neural processes (NPs) learn stochastic processes and predict the distribution of target output adaptively conditioned on a context set of observed input-output pairs. Furthermore, Attentive Neural Process (ANP) improved the prediction accuracy of NPs by incorporating attention mechanism among contexts and targets. In a number of real-world applications such as robotics, finance, speech, and biology, it is critical to learn the temporal order and recurrent structure from sequential data. However, the capability of NPs capturing these properties is limited due to its permutation invariance instinct. In this paper, we proposed the Recurrent Attentive Neural Process (RANP), or alternatively, Attentive Neural Process-RecurrentNeural Network(ANP-RNN), in which the ANP is incorporated into a recurrent neural network. The proposed model encapsulates both the inductive biases of recurrent neural networks and also the strength of NPs for modelling uncertainty. We demonstrate that RANP can effectively model sequential data and outperforms NPs and LSTMs remarkably in a 1D regression toy example as well as autonomous-driving applications.

Chengyuan Zhang, Jiacheng Zhu, Wenshuo Wang et al.

Semantic learning and understanding of multi-vehicle interaction patterns in a cluttered driving environment are essential but challenging for autonomous vehicles to make proper decisions. This paper presents a general framework to gain insights into intricate multi-vehicle interaction patterns from bird's-eye view traffic videos. We adopt a Gaussian velocity field to describe the time-varying multi-vehicle interaction behaviors and then use deep autoencoders to learn associated latent representations for each temporal frame. Then, we utilize a hidden semi-Markov model with a hierarchical Dirichlet process as a prior to segment these sequential representations into granular components, also called traffic primitives, corresponding to interaction patterns. Experimental results demonstrate that our proposed framework can extract traffic primitives from videos, thus providing a semantic way to analyze multi-vehicle interaction patterns, even for cluttered driving scenarios that are far messier than human beings can cope with.

Yaohui Guo, Vinay Varma Kalidindi, Mansur Arief et al.

Autonomous vehicles are expected to navigate in complex traffic scenarios with multiple surrounding vehicles. The correlations between road users vary over time, the degree of which, in theory, could be infinitely large, thus posing a great challenge in modeling and predicting the driving environment. In this paper, we propose a method to model multi-vehicle interactions using a stochastic vector field model and apply non-parametric Bayesian learning to extract the underlying motion patterns from a large quantity of naturalistic traffic data. We then use this model to reproduce the high-dimensional driving scenarios in a finitely tractable form. We use a Gaussian process to model multi-vehicle motion, and a Dirichlet process to assign each observation to a specific scenario. We verify the effectiveness of the proposed method on highway and intersection datasets from the NGSIM project, in which complex multi-vehicle interactions are prevalent. The results show that the proposed method can capture motion patterns from both settings, without imposing heroic prior, and hence demonstrate the potential application for a wide array of traffic situations. The proposed modeling method could enable simulation platforms and other testing methods designed for autonomous vehicle evaluation, to easily model and generate traffic scenarios emulating large scale driving data.

Songlin Xu, Jiacheng Zhu

In order to operate safely on the road, autonomous vehicles need not only to be able to identify objects in front of them, but also to be able to estimate the risk level of the object in front of the vehicle automatically. It is obvious that different objects have different levels of danger to autonomous vehicles. An evaluation system is needed to automatically determine the danger level of the object for the autonomous vehicle. It would be too subjective and incomplete if the system were completely defined by humans. Based on this, we propose a framework based on nonparametric Bayesian learning method -- a sticky hierarchical Dirichlet process hidden Markov model(sticky HDP-HMM), and discover the relationship between driving scenarios and driving styles. We use the analysis of driving styles of autonomous vehicles to reflect the risk levels of driving scenarios to the vehicles. In this framework, we firstly use sticky HDP-HMM to extract driving styles from the dataset and get different clusters, then an evaluation system is proposed to evaluate and rank the urgency levels of the clusters. Finally, we map the driving scenarios to the ranking results and thus get clusters of driving scenarios in different risk levels. More importantly, we find the relationship between driving scenarios and driving styles. The experiment shows that our framework can cluster and rank driving styles of different urgency levels and find the relationship between driving scenarios and driving styles and the conclusions also fit people's common sense when driving. Furthermore, this framework can be used for autonomous vehicles to estimate risk levels of driving scenarios and help them make precise and safe decisions.

Jiacheng Zhu, Wenshuo Wang, Ding Zhao

A multitude of publicly-available driving datasets and data platforms have been raised for autonomous vehicles (AV). However, the heterogeneities of databases in size, structure and driving context make existing datasets practically ineffective due to a lack of uniform frameworks and searchable indexes. In order to overcome these limitations on existing public datasets, this paper proposes a data unification framework based on traffic primitives with ability to automatically unify and label heterogeneous traffic data. This is achieved by two steps: 1) Carefully arrange raw multidimensional time series driving data into a relational database and then 2) automatically extract labeled and indexed traffic primitives from traffic data through a Bayesian nonparametric learning method. Finally, we evaluate the effectiveness of our developed framework using the collected real vehicle data.