Jingwen Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Jingwen Wang has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Stefana Parascho
EPFL
3 shared publications - Gabriel Vallat
EPFL, Lausanne, Switzerland
2 shared publications - Maryam Kamgarpour
Department of Electrical Engineering, EPFL, Lausanne, Switzerland
2 shared publications - Anna Maddux
Department of Electrical Engineering, EPFL, Lausanne, Switzerland
1 shared publication - Fan Shi
National University of Singapore, Singapore, Singapore
1 shared publication - Johannes Kirschner
University of Alberta
1 shared publication
Latest Publications
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (12) , 14708-14715
Learning to Build: Autonomous Robotic Assembly of Stable Structures Without Predefined Plans
- 1 December 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (12) , 14708-14715
Learning to Build: Autonomous Robotic Assembly of Stable Structures Without Predefined Plans
- Open Access
- 27 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 44 (4) , 1-16
Learning to Assemble with Alternative Plans
We present a reinforcement learning framework for constructing assemblies composed of rigid parts, which are commonly seen in many historical masonry buildings and bridges. Traditional construction methods for such structures often depend on dense scaffolding to stabilize their intermediate assembly steps, making the process both labor-intensive and time-consuming. This work utilizes multiple robots to collaboratively assemble structures, offering temporary support by holding parts in place without additional scaffolding. Precomputing the robotic assembly process to ensure structural stability involves a time-consuming offline process due to the combinatorial nature of its search space. However, the precomputed assembly plans may get disrupted during real-world execution due to unforeseen changes, such as setup modifications or delays in part delivery. Recomputing these plans using traditional offline methods results in significant project delays. Therefore, we propose a reinforcement learning-based approach in which a neural network is trained to efficiently generate alternative assembly plans for a given structure online, enabling adaptation to external changes. To enable effective and efficient training, we introduce three key innovations: a GPU-based stability simulator for parallelizing simulations, a novel curriculum-based training scheme to address sparse rewards during training, and a new graph neural network architecture for efficiently encoding assembly geometry. We validate our approach by training reinforcement learning agents on various assemblies and evaluating their performance on unseen assembly tasks. Furthermore, we demonstrate the effectiveness of our framework in planning multi-robot assembly processes, effectively handling disruptions in both simulation and physical environments.
- 27 July 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 44 (4) , 1-16
Open AccessLearning to Assemble with Alternative Plans
We present a reinforcement learning framework for constructing assemblies composed of rigid parts, which are commonly seen in many historical masonry buildings and bridges. Traditional construction methods for such structures often depend on dense scaffolding to stabilize their intermediate assembly steps, making the process both labor-intensive and time-consuming. This work utilizes multiple robots to collaboratively assemble structures, offering temporary support by holding parts in place without additional scaffolding. Precomputing the robotic assembly process to ensure structural stability involves a time-consuming offline process due to the combinatorial nature of its search space. However, the precomputed assembly plans may get disrupted during real-world execution due to unforeseen changes, such as setup modifications or delays in part delivery. Recomputing these plans using traditional offline methods results in significant project delays. Therefore, we propose a reinforcement learning-based approach in which a neural network is trained to efficiently generate alternative assembly plans for a given structure online, enabling adaptation to external changes. To enable effective and efficient training, we introduce three key innovations: a GPU-based stability simulator for parallelizing simulations, a novel curriculum-based training scheme to address sparse rewards during training, and a new graph neural network architecture for efficiently encoding assembly geometry. We validate our approach by training reinforcement learning agents on various assemblies and evaluating their performance on unseen assembly tasks. Furthermore, we demonstrate the effectiveness of our framework in planning multi-robot assembly processes, effectively handling disruptions in both simulation and physical environments.
- Open Access
- 8 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Reinforcement learning for scaffold-free construction of spanning structures
In construction robotics, a conventional design-to-fabrication workflow starts with designing a structure, followed by task and robotic motion planning, and ultimately, fabrication. However, this approach can prove unsuccessful, as we may only discover the infeasibility of a design at the final stages of the process. This can result in rework and a considerable waste of time and resources. To overcome this challenge, we propose a design method based on reinforcement learning (RL) where the agent makes decisions at every step of the sequential assembly of the structure while considering assembly’s stability. In this way, we take the construction constraints into consideration at the design stage. The research particularly focuses on the design of spanning structures that multiple robot arms can construct without the need for scaffolding. A series of experiments were conducted using both a centralized and a decentralized learning setup. Our results show that while the decentralized setup was successful in constructing smaller structures, only the centralized setup allowed active collaboration between robot arms, resulting in structures with larger spans. To validate our approach, we fabricated two of the designed structures with two collaborating robot arms, which confirmed the feasibility of these designs. In summary, the proposed method opens exciting possibilities for generating innovative designs that push the boundaries of architectural creativity while simultaneously fulfilling fabrication-related constraints.
- 8 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Open AccessReinforcement learning for scaffold-free construction of spanning structures
In construction robotics, a conventional design-to-fabrication workflow starts with designing a structure, followed by task and robotic motion planning, and ultimately, fabrication. However, this approach can prove unsuccessful, as we may only discover the infeasibility of a design at the final stages of the process. This can result in rework and a considerable waste of time and resources. To overcome this challenge, we propose a design method based on reinforcement learning (RL) where the agent makes decisions at every step of the sequential assembly of the structure while considering assembly’s stability. In this way, we take the construction constraints into consideration at the design stage. The research particularly focuses on the design of spanning structures that multiple robot arms can construct without the need for scaffolding. A series of experiments were conducted using both a centralized and a decentralized learning setup. Our results show that while the decentralized setup was successful in constructing smaller structures, only the centralized setup allowed active collaboration between robot arms, resulting in structures with larger spans. To validate our approach, we fabricated two of the designed structures with two collaborating robot arms, which confirmed the feasibility of these designs. In summary, the proposed method opens exciting possibilities for generating innovative designs that push the boundaries of architectural creativity while simultaneously fulfilling fabrication-related constraints.