Capturing Car-Following Behaviors by Deep Learning

Wang, Xiao; Jiang, Rui; Li, Li; Lin, Yilun; Zheng, Xinhu; Wang, Fei‐Yue

doi:10.1109/tits.2017.2706963

Cited by 252 publications

(142 citation statements)

References 55 publications

Supporting

Mentioning

140

Contrasting

Order By: Relevance

“…Chong et al (30) illustrated that it is possible to predict acceleration accurately using neural networks with only one hidden layer, and Khodayari et al (31) instantaneous reaction time (RT) delay. In contrast to these conventional neural network-based models, recent studies have begun to model car-following behavior with recurrent neural networks (RNN) (32), and with more than one hidden layer, i.e., deep neural networks (8).…”

Section: Data-driven Car-following Modelsmentioning

confidence: 99%

“…The critic network with weights θ Q approximates the action-value function Q(s,a|θ Q ). The actor network with weights θ µ explicitly 8 represents the agent's current policy µ(s|θ µ ) for the Q-function, which maps from a state of the environment s to an action a. To enable stable and robust learning, DDPG deploys experience replay and target network, as in DQN:…”

Section: Deep Deterministic Policy Gradientmentioning

confidence: 99%

See 1 more Smart Citation

Human-like autonomous car-following model with deep reinforcement learning

Zhu

Wang

2018

Transportation Research Part C: Emerging Technologies

390

155

View full text Add to dashboard Cite

This study proposes a framework for human-like autonomous car-following planning based on deep reinforcement learning (deep RL). Historical driving data are fed into a simulation environment where an RL agent learns from trial and error interactions based on a reward function that signals how much the agent deviates from the empirical data. Through these interactions, an optimal policy, or car-following model that maps in a human-like way from speed, relative speed between a lead and following vehicle, and inter-vehicle spacing to acceleration of a following vehicle is finally obtained. The model can be continuously updated when more data are fed in. Two thousand car-following periods extracted from the 2015 Shanghai Naturalistic Driving Study were used to train the model and compare its performance with that of traditional and recent datadriven car-following models. As shown by this study's results, a deep deterministic policy gradient car-following model that uses disparity between simulated and observed speed as the reward function and considers a reaction delay of 1s, denoted as DDPGvRT, can reproduce human-like car-following behavior with higher accuracy than traditional and recent data-driven car-following models. Specifically, the DDPGvRT model has a spacing validation error of 18% and speed validation error of 5%, which are less than those of other models, including the intelligent driver model, models based on locally weighted regression, and conventional neural network-based models. Moreover, the DDPGvRT demonstrates good capability of generalization to various driving situations and can adapt to different drivers by continuously learning. This study demonstrates that reinforcement learning methodology can offer insight into driver behavior and can contribute to the development of human-like autonomous driving algorithms and traffic-flow models.

show abstract

Section: Data-driven Car-following Modelsmentioning

confidence: 99%

Section: Deep Deterministic Policy Gradientmentioning

confidence: 99%

Human-like autonomous car-following model with deep reinforcement learning

Zhu

Wang

2018

Transportation Research Part C: Emerging Technologies

390

155

View full text Add to dashboard Cite

show abstract

“…RNNs have been widely used to capture temporal dependency in sequential data [35], [36]. Representative models include GRUs [37] and long short term memory (LSTM) [38], and they have been adopted successfully to various tasks such as video representation [39], image captioning [40] and carfollowing modeling [41]. LSTM and GRUs, typically using fully-connected layers, do not maintain spatial information and require a lot of network parameters.…”

Section: B Recurrent Modelsmentioning

confidence: 99%

Temporally Consistent Depth Prediction With Flow-Guided Memory Units

Eom

Park

Ham

2020

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

Predicting depth from a monocular video sequence is an important task for autonomous driving. Although it has advanced considerably in the past few years, recent methods based on convolutional neural networks (CNNs) discard temporal coherence in the video sequence and estimate depth independently for each frame, which often leads to undesired inconsistent results over time. To address this problem, we propose to memorize temporal consistency in the video sequence, and leverage it for the task of depth prediction. To this end, we introduce a two-stream CNN with a flow-guided memory module, where each stream encodes visual and temporal features, respectively. The memory module, implemented using convolutional gated recurrent units (ConvGRUs), inputs visual and temporal features sequentially together with optical flow tailored to our task. It memorizes trajectories of individual features selectively and propagates spatial information over time, enforcing a long-term temporal consistency to prediction results. We evaluate our method on the KITTI benchmark dataset in terms of depth prediction accuracy, temporal consistency and runtime, and achieve a new state of the art. We also provide an extensive experimental analysis, clearly demonstrating the effectiveness of our approach to memorizing temporal consistency for depth prediction.

show abstract

“…To solve this problem, various modifications have been made on RNNs; the most effective method is to introduce various gating mechanisms such as the long short-term memory (LSTM) [18,19] and the gated recurrent unit (GRU) [20] networks. Based on the previously described studies, Wang et al [21] proposed the use of GRU to model CF behaviour and embed the driver's memory effect in the model, which used the speed of the following car, the relative speed of the two cars, and the distance between the two cars observed in the last several time intervals as inputs and the estimated speed of the following car at the next time point as the output. e test results showed that the proposed model has higher simulation accuracy than the existing CF models and provides a new concept for the study of traffic flow theory and simulation.…”

Section: Introductionmentioning

confidence: 99%

The Driver Time Memory Car-Following Model Simulating in Apollo Platform with GRU and Real Road Traffic Data

Fei

Hei

et al. 2020

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Car following is the most common phenomenon in single-lane traffic. The accuracy of acceleration prediction can be effectively improved by the driver’s memory in car-following behaviour. In addition, the Apollo autonomous driving platform launched by Baidu Inc. provides fast test vehicle following vehicle models. Therefore, this paper proposes a car-following model (CFDT) with driver time memory based on real-world traffic data. The CFDT model is firstly constructed by embedded gantry control unit storage capacity (GRU assisted) network. Secondly, the NGSIM dataset will be used to obtain the tracking data of small vehicles with similar driving behaviours from the common real road vehicle driving tracks for data preprocessing according to the response time of drivers. Then, the model is calibrated to obtain the driver’s driving memory and the optimal parameters of the model and structure. Finally, the Apollo simulation platform with high-speed automatic driving technology is used for 3D visualization interface verification. Comparative experiments on vehicle tracking characteristics show that the CFDT model is effective and robust, which improves the simulation accuracy. Meanwhile, the model is tested and validated using the Apollo simulation platform to ensure accuracy and utility of the model.

show abstract

Capturing Car-Following Behaviors by Deep Learning

Cited by 252 publications

References 55 publications

Human-like autonomous car-following model with deep reinforcement learning

Human-like autonomous car-following model with deep reinforcement learning

Temporally Consistent Depth Prediction With Flow-Guided Memory Units

The Driver Time Memory Car-Following Model Simulating in Apollo Platform with GRU and Real Road Traffic Data

Contact Info

Product

Resources

About