Lian Hou

Lian Hou

Auto-generated profile, based on publicly available information

Lian Hou

Institute of Industrial Science, The University of Tokyo, Tokyo, Japan

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 August 2022 in IEEE Transactions on Intelligent Transportation Systems
Verified
Journal Article
Closed access
LHLian Hou
Lian Hou
SLShengbo Li
Shengbo Li
BYBo Yang
Bo Yang
ZWZheng Wang
Zheng Wang
KNKimihiko Nakano
Kimihiko Nakano
Fast and accurate long-term trajectory prediction of surrounding vehicles (SVs) is critical to autonomous driving systems. In high-density traffic flows, strongly correlated vehicle behaviors require considering the interactions among multiple SVs when predicting their future trajectories. However, existing interactive prediction methods, most based on Long Short-Term Memory (LSTM), are suffering from slow prediction because they analyze SVs one by one and analyze trajectory sequence node by node. This paper presents a fast interactive trajectory prediction method called Structural Transformer which learns both spatial and temporal dependencies among multiple SVs in parallel. Specifically, our model first removes the internal states and loops of LSTM and replaces with a weighted self-reference mapping to realize parallel computation. Then, it embeds the relative spatial information of multiple SVs into trajectory states and reorganizes the self-reference mapping with neighbor-only interaction masks to achieve interactive prediction. Results on the NGSIM dataset show satisfyingly speed and accuracy performance on long-term trajectory prediction of multiple SVs. The longitudinal and lateral errors are reduced to 2.67m and 0.25m over 5s time horizon. The computational time of each step is only 12ms on a 2080ti GPU, which is over 4 times faster than the Structural LSTM.

8 August 2022 in IEEE Transactions on Intelligent Vehicles
Verified
Journal Article
Closed access
LHLian Hou
Lian Hou
SLShengbo Li
Shengbo Li
BYBo Yang
Bo Yang
ZWZheng Wang
Zheng Wang
KNKimihiko Nakano
Kimihiko Nakano
Varying numbers and types of vehicles, various road structures and traffic rules bring difficulties to an autonomous vehicle driving in highway traffic scenarios. It is important to simultaneously consider all these elements in an integrated framework when predicting the future trajectories of surrounding vehicles. This paper presents a unified graphical representation method for dynamic traffic scenarios based on integrating not only the constraints from vehicles but also the collision risk implied behind road structures and traffic rules. Different from previous studies which ignores road structures and traffic rules or separately represent them in a qualitative way, this method can make better use of the influences of these environment elements in a quantitative way to improve trajectory prediction of surrounding vehicles in highway scenarios. Specifically, this method first assesses the permissibility and passability of different types of roads and vehicles, and then defines their collision risk level based on traffic rules like speed limit or priority signs. Secondly, the bird-view scenarios are rasterized and rendered as color-scaled feature maps based on this risk level. Finally, multiple feature maps with various spatial and temporal perspective field are sequentially fed to a convolutional long short-term memory (LSTM) network to predict trajectories for surrounding vehicles. We implement the model on NGSIM dataset and achieve satisfying prediction accuracy, for example, the error over 5s time prediction horizon can be reduced to less than 2.78m in longitudinal and 0.26m in lateral direction.