Learning to Predict Lane Changes in Highway Scenarios Using Dynamic Filters On a Generic Traffic Representation

Mänttäri, Joonatan; Folkesson, John; Ward, Erik

doi:10.1109/ivs.2018.8500426

Cited by 10 publications

(17 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, and based on their success in other domains, more advanced methods utilizing Recurrent Neural Networks (RNN) [14], deep learning [15]- [17], and reinforcement learning have arisen as well [18]. Convolutional Neural Networks (CNN) are also used for LC detection using image prediction [19], [20]. Learning based approaches require high-quality motion datasets containing interactive real-world driving scenarios for training and testing.…”

Section: A Maneuver Classificationmentioning

confidence: 99%

Artificial Intelligence for Vehicle Behavior Anticipation: Hybrid Approach Based on Maneuver Classification and Trajectory Prediction

et al. 2020

View full text Add to dashboard Cite

Innovative technologies and naturalistic driving data sources provide a great potential to develop reliable autonomous driving systems. Understanding the behaviors of surrounding vehicles is essential for improving safety and mobility of autonomous vehicles. Onboard sensors like Radar, Lidar and Camera are able to track surrounding vehicles motion and to get different features like position, velocity and yaw. This paper proposes a hybrid approach to integrate maneuver classification using neural networks and trajectory prediction using Long Short-term Memory (LSTM) networks to get the future positions of adjacent vehicles. In this study we use the Next Generation Simulation (NGSIM) public dataset that provides a real driving data. The proposed approach is validated experimentally using VEDECOM demonstrator data. The results demonstrate that the proposed approach is able to predict driver intention to change lanes on average 2.2 seconds in advance. The Root Mean Square (RMS) errors of lateral and longitudinal positions are 0.30 m and 3.1 m respectively. The results demonstrate a high performance compared to various existing methods.

show abstract

Section: A Maneuver Classificationmentioning

confidence: 99%

Artificial Intelligence for Vehicle Behavior Anticipation: Hybrid Approach Based on Maneuver Classification and Trajectory Prediction

et al. 2020

View full text Add to dashboard Cite

show abstract

“…In the proposed method, the goal is to produce a prediction system that incorporates uncertainty and takes all available interactions between an arbitrary number of traffic agents into account, while keeping the complexity low. To allow the system to learn the effect of arbitrary interactions on vehicles, the generic scene representation previously proposed in [8] is adopted, explained in detail in section II. This enables the method to learn interactions directly from data without the need to specify explicit models for vehicle dynamics or interactions.…”

Section: A Related Workmentioning

confidence: 99%

“…3. This method is based upon the previously proposed framework in [8], which produces a prediction for the next frame of the traffic sequence by learning Dynamic Filters [11] that operate on the same input sequence of 10 frames as the CVAE method. After having produced one prediction, the oldest frame in the input sequence is discarded and the prediction is used as the newest frame in the input sequence.…”

Section: B Prediction Generationmentioning

confidence: 99%

“…Fig. 3: System architecture of the MC Dropout prediction method, derived from previous work in [8] by adding dropout in the fully connected layers and sampling the resulting predictions multiple times.…”

Section: Trainingmentioning

confidence: 99%

“…The method proposed in this paper utilizes a single Deep Learning-based module that is capable of learning complex interactions between agents, by leveraging a generic image representation of the traffic scenario and capturing spatiotemporal characteristics through recurrent convolutional operations, previously shown to be effective in predicting lane change maneuvers in highway scenarios [8]. In order to incorporate uncertainty two approaches are evaluated, Monte Carlo Dropout and using a Conditional Variational Auto Encoder.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Incorporating Uncertainty in Predicting Vehicle Maneuvers at Intersections With Complex Interactions

Mänttäri¹,

Folkesson²

2019

2019 IEEE Intelligent Vehicles Symposium (IV)

Self Cite

View full text Add to dashboard Cite

Highly automated driving systems are required to make robust decisions in many complex driving environments, such as urban intersections with high traffic. In order to make as informed and safe decisions as possible, it is necessary for the system to be able to predict the future maneuvers and positions of other traffic agents, as well as to provide information about the uncertainty in the prediction to the decision making module. While Bayesian approaches are a natural way of modeling uncertainty, recently deep learning-based methods have emerged to address this need as well. However, balancing the computational and system complexity, while also taking into account agent interactions and uncertainties, remains a difficult task. The work presented in this paper proposes a method of producing predictions of other traffic agents' trajectories in intersections with a singular Deep Learning module, while incorporating uncertainty and the interactions between traffic participants. The accuracy of the generated predictions is tested on a simulated intersection with a high level of interaction between agents, and different methods of incorporating uncertainty are compared. Preliminary results show that the CVAE-based method produces qualitatively and quantitatively better measurements of uncertainty and manage to more accurately assign probability to the future occupied space of traffic agents.

show abstract

Environmental model extension for lane change prediction with neural networks

et al. 2019

View full text Add to dashboard Cite

Learning to Predict Lane Changes in Highway Scenarios Using Dynamic Filters On a Generic Traffic Representation

Cited by 10 publications

References 12 publications

Artificial Intelligence for Vehicle Behavior Anticipation: Hybrid Approach Based on Maneuver Classification and Trajectory Prediction

Artificial Intelligence for Vehicle Behavior Anticipation: Hybrid Approach Based on Maneuver Classification and Trajectory Prediction

Incorporating Uncertainty in Predicting Vehicle Maneuvers at Intersections With Complex Interactions

Environmental model extension for lane change prediction with neural networks

Contact Info

Product

Resources

About