Wenxuan Ao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Wenxuan Ao has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Yong Li
Tsinghua University
3 shared publications - Depeng Jin
Beijing National Research Center for Information Science and Technology (BNRist), Department of Electronic Engineering, Tsinghua University, Beijing, China
2 shared publications - Chao Yu
Tsinghua University
1 shared publication - Guozhen Zhang
TsingRoc.ai, Beijing, China
1 shared publication - Huazhong Yang
Tsinghua University
1 shared publication - Jian Yuan
Tsinghua University
1 shared publication
Latest Publications
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 4036-4044
CityLight: A Neighborhood-inclusive Universal Model for Coordinated City-scale Traffic Signal Control
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 4036-4044
Open AccessCityLight: A Neighborhood-inclusive Universal Model for Coordinated City-scale Traffic Signal Control
- Open Access
- 13 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 23-32
A City-level High-performance Spatio-temporal Mobility Simulation System
Urban mobility simulation refers to simulating human fine-grained spatio-temporal mobility and activity behaviors in cities, which facilitates the measurement of traffic operations, assesses the impacts of transportation on other areas of the city like environment, and supports the designation of simulation-driven mobility-related sustainable policies. It is becoming one of the important tools in the development of sustainable cities. However, the main challenge currently restricting the mobility simulation and its applications is poor performance when dealing with city-scale million or even ten million people simulation. In mobility simulation, agents like people and vehicles need to take actions based on the previous step's states of other agents nearby, which reflects the spatio-temporal dependencies among agents. Facing city-scale scenarios, the spatio-temporal dependencies become the main barrier to achieving efficient parallel computation acceleration. To alleviate the impact of spatio-temporal dependencies on parallel acceleration, we propose a city-level high-performance spatio-temporal mobility simulation system designed with a two-stage parallel process based on read/write separation and a parallel-friendly indexing subsystem. The two-stage parallel process optimizes cross-step state read/write processes among agents by reorganizing all states into three categories (public read-only, public write-only, and private) and introducing a two-stage control flow design to divide the entire data flow into two easily parallelized groups. The indexing subsystem optimizes spatial belonging relationship maintenance and relative location queries through parallel-friendly data structure selection with addition, deletion, and query process design. We implement the whole system on both CPU and GPU to adapt to different hardware environments. We also conduct extensive experiments and build use cases to demonstrate that the system achieves the expected results in terms of performance and can support innovative applications about sustainable mobility research. The experiments show that our proposed system achieves a computational speedup of 278.77 times the wall clock time, i.e. 3.59 milliseconds per step, in a city-level simulation with nearly 1 million people simultaneously.
- 13 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 23-32
Open AccessA City-level High-performance Spatio-temporal Mobility Simulation System
Urban mobility simulation refers to simulating human fine-grained spatio-temporal mobility and activity behaviors in cities, which facilitates the measurement of traffic operations, assesses the impacts of transportation on other areas of the city like environment, and supports the designation of simulation-driven mobility-related sustainable policies. It is becoming one of the important tools in the development of sustainable cities. However, the main challenge currently restricting the mobility simulation and its applications is poor performance when dealing with city-scale million or even ten million people simulation. In mobility simulation, agents like people and vehicles need to take actions based on the previous step's states of other agents nearby, which reflects the spatio-temporal dependencies among agents. Facing city-scale scenarios, the spatio-temporal dependencies become the main barrier to achieving efficient parallel computation acceleration. To alleviate the impact of spatio-temporal dependencies on parallel acceleration, we propose a city-level high-performance spatio-temporal mobility simulation system designed with a two-stage parallel process based on read/write separation and a parallel-friendly indexing subsystem. The two-stage parallel process optimizes cross-step state read/write processes among agents by reorganizing all states into three categories (public read-only, public write-only, and private) and introducing a two-stage control flow design to divide the entire data flow into two easily parallelized groups. The indexing subsystem optimizes spatial belonging relationship maintenance and relative location queries through parallel-friendly data structure selection with addition, deletion, and query process design. We implement the whole system on both CPU and GPU to adapt to different hardware environments. We also conduct extensive experiments and build use cases to demonstrate that the system achieves the expected results in terms of performance and can support innovative applications about sustainable mobility research. The experiments show that our proposed system achieves a computational speedup of 278.77 times the wall clock time, i.e. 3.59 milliseconds per step, in a city-level simulation with nearly 1 million people simultaneously.
- Open Access
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 3713-3723
Learning to Solve Grouped 2D Bin Packing Problems in the Manufacturing Industry
The two-dimensional bin packing problem (2DBP) is a critical optimization problem in the furniture production and glass cutting industries, where the objective is to cut smaller-sized items from a minimum number of large standard-sized raw materials. In practice, factories manufacture hundreds of customer orders (sets of items) every day, and to relieve pressure in management, a common practice is to group the orders into batches for production, ensuring that items from one order are in the same batch instead of scattered across the production line. In this work, we formulate this problem as the grouped 2D bin packing problem, a bi-level problem where the upper level partitions orders into groups and the lower level solves 2DBP for items in each group. The main challenges are (1) the coupled optimization of upper and lower levels and (2) the high computational efficiency required for practical application. To tackle these challenges, we propose an iteration-based hierarchical reinforcement learning framework, which can learn to solve the optimization problem in a data-driven way and provide fast online performance after offline training. Extensive experiments demonstrate that our method not only achieves the best performance compared to all baselines but is also robust to changes in dataset distribution and problem constraints. Finally, we deployed our method in the ARROW Home factory in China, resulting in a 4.1% reduction in raw material costs. We have released the source code and datasets to facilitate future research.
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 3713-3723
Open AccessLearning to Solve Grouped 2D Bin Packing Problems in the Manufacturing Industry
The two-dimensional bin packing problem (2DBP) is a critical optimization problem in the furniture production and glass cutting industries, where the objective is to cut smaller-sized items from a minimum number of large standard-sized raw materials. In practice, factories manufacture hundreds of customer orders (sets of items) every day, and to relieve pressure in management, a common practice is to group the orders into batches for production, ensuring that items from one order are in the same batch instead of scattered across the production line. In this work, we formulate this problem as the grouped 2D bin packing problem, a bi-level problem where the upper level partitions orders into groups and the lower level solves 2DBP for items in each group. The main challenges are (1) the coupled optimization of upper and lower levels and (2) the high computational efficiency required for practical application. To tackle these challenges, we propose an iteration-based hierarchical reinforcement learning framework, which can learn to solve the optimization problem in a data-driven way and provide fast online performance after offline training. Extensive experiments demonstrate that our method not only achieves the best performance compared to all baselines but is also robust to changes in dataset distribution and problem constraints. Finally, we deployed our method in the ARROW Home factory in China, resulting in a 4.1% reduction in raw material costs. We have released the source code and datasets to facilitate future research.