Search

Your search keyword '"Wu, Chenyang"' showing total 420 results
Start Over Author "Wu, Chenyang"
420 results on '"Wu, Chenyang"'

1. MM-Gaussian: 3D Gaussian-based Multi-modal Fusion for Localization and Reconstruction in Unbounded Scenes

Author

Wu, Chenyang, Duan, Yifan, Zhang, Xinran, Sheng, Yu, Ji, Jianmin, and Zhang, Yanyong

Subjects
Computer Science - Robotics, Computer Science - Computer Vision and Pattern Recognition
Abstract

Localization and mapping are critical tasks for various applications such as autonomous vehicles and robotics. The challenges posed by outdoor environments present particular complexities due to their unbounded characteristics. In this work, we present MM-Gaussian, a LiDAR-camera multi-modal fusion system for localization and mapping in unbounded scenes. Our approach is inspired by the recently developed 3D Gaussians, which demonstrate remarkable capabilities in achieving high rendering quality and fast rendering speed. Specifically, our system fully utilizes the geometric structure information provided by solid-state LiDAR to address the problem of inaccurate depth encountered when relying solely on visual solutions in unbounded, outdoor scenarios. Additionally, we utilize 3D Gaussian point clouds, with the assistance of pixel-level gradient descent, to fully exploit the color information in photos, thereby achieving realistic rendering effects. To further bolster the robustness of our system, we designed a relocalization module, which assists in returning to the correct trajectory in the event of a localization failure. Experiments conducted in multiple scenarios demonstrate the effectiveness of our method., Comment: 7 pages, 5 figures

Published
2024

2. Reinforced In-Context Black-Box Optimization

Author

Song, Lei, Gao, Chenxiao, Xue, Ke, Wu, Chenyang, Li, Dong, Hao, Jianye, Zhang, Zongzhang, and Qian, Chao

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Neural and Evolutionary Computing
Abstract

Black-Box Optimization (BBO) has found successful applications in many fields of science and engineering. Recently, there has been a growing interest in meta-learning particular components of BBO algorithms to speed up optimization and get rid of tedious hand-crafted heuristics. As an extension, learning the entire algorithm from data requires the least labor from experts and can provide the most flexibility. In this paper, we propose RIBBO, a method to reinforce-learn a BBO algorithm from offline data in an end-to-end fashion. RIBBO employs expressive sequence models to learn the optimization histories produced by multiple behavior algorithms and tasks, leveraging the in-context learning ability of large models to extract task information and make decisions accordingly. Central to our method is to augment the optimization histories with \textit{regret-to-go} tokens, which are designed to represent the performance of an algorithm based on cumulative regret over the future part of the histories. The integration of regret-to-go tokens enables RIBBO to automatically generate sequences of query points that satisfy the user-desired regret, which is verified by its universally good empirical performance on diverse problems, including BBO benchmark functions, hyper-parameter optimization and robot control problems.

Published
2024

4. ACT: Empowering Decision Transformer with Dynamic Programming via Advantage Conditioning

Author

Gao, Chen-Xiao, Wu, Chenyang, Cao, Mingjun, Kong, Rui, Zhang, Zongzhang, and Yu, Yang

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Decision Transformer (DT), which employs expressive sequence modeling techniques to perform action generation, has emerged as a promising approach to offline policy optimization. However, DT generates actions conditioned on a desired future return, which is known to bear some weaknesses such as the susceptibility to environmental stochasticity. To overcome DT's weaknesses, we propose to empower DT with dynamic programming. Our method comprises three steps. First, we employ in-sample value iteration to obtain approximated value functions, which involves dynamic programming over the MDP structure. Second, we evaluate action quality in context with estimated advantages. We introduce two types of advantage estimators, IAE and GAE, which are suitable for different tasks. Third, we train an Advantage-Conditioned Transformer (ACT) to generate actions conditioned on the estimated advantages. Finally, during testing, ACT generates actions conditioned on a desired advantage. Our evaluation results validate that, by leveraging the power of dynamic programming, ACT demonstrates effective trajectory stitching and robust action generation in spite of the environmental stochasticity, outperforming baseline methods across various benchmarks. Additionally, we conduct an in-depth analysis of ACT's various design choices through ablation studies. Our code is available at https://github.com/LAMDA-RL/ACT., Comment: Accepted by AAAI 2024

Published
2023

37. Carsharing : user behaviour, system management and pricing under uncertainty

Author

Wu, Chenyang, Sivakumar, Aruna, and Le Vine, Scott

Subjects
388.4
Abstract

Free-floating carsharing systems (FFCS) are characterised by the volatile spatial-temporal distribution of fleet and user demand, as well as users’ heterogenous price sensitivity and spatial-temporal flexibility. Hence, it is possible to apply dynamic pricing-based revenue management in a FFCS system. By dynamically updating prices on different origin-destination pairs according to the realtime fleet and user demand distribution, operators can redistribute user demand as well as increase their own revenue. In addition to dynamic pricing, advance reservation also has the potential to reduce the volatility of FFCS system and make the carsharing service more profitable. By having advance reservation, the operator has better knowledge of future user demand and can prepare in advance for possible demand peaks. This increase in service reliability can better serve the current FFCS user and attract potential new FFCS users. These two approaches have been investigated by some of the literature in the carsharing research, especially dynamic pricing. The success of both approaches requires the participation of users, however, and we are still unclear about users’ preferences in respect to the two approaches. Without a clear understanding of user behaviour, it is too early to assume that the two approaches can help the operator in revenue management. Dynamic pricing, especially, adds an additional dimension of uncertainty to the volatile free-floating carsharing system, which may raise users’ aversion and not help the operator increase the revenue. This thesis contributes to the wider literature by bridging user behaviour modelling and carsharing system operational management. We estimate users’ preference for/against dynamic pricing and advance reservation, and develop a decision support tool that helps the carsharing operator to maximise revenue while taking user behaviour into consideration. We design a stated choice (SC) survey to collect carsharing users’ responses to two interacting dimensions of uncertainties. For choice-behaviour under multiple dimensions of uncertainty, we propose two levels of transformation (attribute- and utility-level transformation). We estimate carsharing user behaviour under uncertainty through a discrete choice model (DCM). The modelling results demonstrate that our SC survey design approach can successfully capture carsharing user behaviour under two interactive dimensions of uncertainty. The modelling results from the two levels of transformation are consistent with each other, and help the carsharing operator understand the user behaviour from different angles. This demonstrates the importance of considering both levels of transformation in risky-choice behaviour modelling with more than one dimension of uncertainty. We also provide a choice-based optimisation framework that considers users’ risky-choice behaviour. Choice behaviour as predicted by the risky-choice models and estimated from the SC survey data feeds into the choice-based optimisation model. We demonstrate the importance of having a correct understanding of users’ risk preference in dynamic pricing-based revenue management by a numerical analysis. The choice-based optimisation framework can serve as a decision support tool for carsharing operators to generate more revenue and better serve the users.

Published
2020
Full Text
View/download PDF

40. Bioinspired Multifunctional Anti-icing Hydrogel

Author

He, Zhiyuan, Wu, Chenyang, Hua, Mutian, Wu, Shuwang, Wu, Dong, Zhu, Xinyuan, Wang, Jianjun, and He, Ximin

Subjects
Affordable and Clean Energy
Abstract

The recent anti-icing strategies in the state of the art mainly focused on three aspects: inhibiting ice nucleation, preventing ice propagation, and decreasing ice adhesion strength. However, it is has proved difficult to prevent ice nucleation and propagation while decreasing adhesion simultaneously, due to their highly distinct, even contradictory design principles. In nature, anti-freeze proteins (AFPs) offer a prime example of multifunctional integrated anti-icing materials that excel in all three key aspects of the anti-icing process simultaneously by tuning the structures and dynamics of interfacial water. Here, inspired by biological AFPs, we successfully created a multifunctional anti-icing material based on polydimethylsiloxane-grafted polyelectrolyte hydrogel that can tackle all three aspects of the anti-icing process simultaneously. The simplicity, mechanical durability, and versatility of these smooth hydrogel surfaces make it a promising option for a wide range of anti-icing applications.

Published
2020

Catalog

Books, media, physical & digital resources
See catalog results