arXiv subjects

Yinan Zheng

Publications and source records attributed to Yinan Zheng.

24 records · Page 2Linked to original sources

arXiv2024

DecisionNCE: Embodied Multimodal Representations via Implicit Preference Learning

Multimodal pretraining is an effective strategy for the trinity of goals of representation learning in autonomous robots: 1) extracting both local and global task progressions; 2) enforcing temporal consistency of visual representation; 3) capturing trajectory-level language grounding. Most existing methods approach these via separate objectives, which often reach sub-optimal solutions. In this paper, we propose a universal unified objective that can simultaneously extract meaningful task progression information from image sequences and seamlessly align them with language instructions. We discover that via implicit preferences, where a visual trajectory inherently aligns better with its corresponding language instruction than mismatched pairs, the popular Bradley-Terry model can transform into representation learning through proper reward reparameterizations. The resulted framework, DecisionNCE, mirrors an InfoNCE-style objective but is distinctively tailored for decision-making tasks, providing an embodied representation learning framework that elegantly extracts both local and global task progression features, with temporal consistency enforced through implicit time contrastive learning, while ensuring trajectory-level instruction grounding via multimodal joint encoding. Evaluation on both simulated and real robots demonstrates that DecisionNCE effectively facilitates diverse downstream policy learning tasks, offering a versatile solution for unified representation and reward learning. Project Page: https://2toinf.github.io/DecisionNCE/

cs.RO↗

arXiv2024

Safe Offline Reinforcement Learning with Feasibility-Guided Diffusion Model

Safe offline RL is a promising way to bypass risky online interactions towards safe policy learning. Most existing methods only enforce soft constraints, i.e., constraining safety violations in expectation below thresholds predetermined. This can lead to potentially unsafe outcomes, thus unacceptable in safety-critical scenarios. An alternative is to enforce the hard constraint of zero violation. However, this can be challenging in offline setting, as it needs to strike the right balance among three highly intricate and correlated aspects: safety constraint satisfaction, reward maximization, and behavior regularization imposed by offline datasets. Interestingly, we discover that via reachability analysis of safe-control theory, the hard safety constraint can be equivalently translated to identifying the largest feasible region given the offline dataset. This seamlessly converts the original trilogy problem to a feasibility-dependent objective, i.e., maximizing reward value within the feasible region while minimizing safety risks in the infeasible region. Inspired by these, we propose FISOR (FeasIbility-guided Safe Offline RL), which allows safety constraint adherence, reward maximization, and offline policy learning to be realized via three decoupled processes, while offering strong safety performance and stability. In FISOR, the optimal policy for the translated optimization problem can be derived in a special form of weighted behavior cloning. Thus, we propose a novel energy-guided diffusion model that does not require training a complicated time-dependent classifier to extract the policy, greatly simplifying the training. We compare FISOR against baselines on DSRL benchmark for safe offline RL. Evaluation results show that FISOR is the only method that can guarantee safety satisfaction in all tasks, while achieving top returns in most tasks.

cs.LG↗

arXiv2023

Offline Multi-Agent Reinforcement Learning with Implicit Global-to-Local Value Regularization

Offline reinforcement learning (RL) has received considerable attention in recent years due to its attractive capability of learning policies from offline datasets without environmental interactions. Despite some success in the single-agent setting, offline multi-agent RL (MARL) remains to be a challenge. The large joint state-action space and the coupled multi-agent behaviors pose extra complexities for offline policy optimization. Most existing offline MARL studies simply apply offline data-related regularizations on individual agents, without fully considering the multi-agent system at the global level. In this work, we present OMIGA, a new offline m ulti-agent RL algorithm with implicit global-to-local v alue regularization. OMIGA provides a principled framework to convert global-level value regularization into equivalent implicit local value regularizations and simultaneously enables in-sample learning, thus elegantly bridging multi-agent value decomposition and policy learning with offline regularizations. Based on comprehensive experiments on the offline multi-agent MuJoCo and StarCraft II micro-management tasks, we show that OMIGA achieves superior performance over the state-of-the-art offline MARL methods in almost all tasks.

cs.LG↗

arXiv2022

Distributed Coverage Control of Multi-Agent Systems in Uncertain Environments using Heat Transfer Equations

This paper addresses the coverage control problem of multi-agent systems in the uncertain environment. With the aid of Voronoi partition, a distributed coverage control formulation of multi-agent system is proposed to complete the workload in uncertain environments. Driven by the gradient of thermal field, each agent is able to move around for clearing the workload on its own subregion. Theoretical analysis is conducted to ensure the completion of workload in finite time. Finally, numerical simulations are carried out to demonstrate the effectiveness and advantages of the proposed coverage control approach as compared to other existing approaches.

math.OC↗

arXiv2020

Topological $R$-pressure and topological pressure of free semigroup actions

In this paper we introduce the definition of topological $r$-pressure of free semigroup actions on compact metric space and provide some properties of it. Through skew-product transformation into a medium, we can obtain the following two main results. 1. We extend the result that the topological pressure is the limit of topological $r$-pressure in\cite{C} to free semigroup actions ($r\to 0$). 2. Let $f_i,$ $i=0, 1, \cdots, m-1$, be homeomorphisms on a compact metric space. For any continuous function, we verify that the topological pressure of $f_0, \cdots, f_{m-1}$ equals the topological pressure of $f_0^{-1}, \cdots, f_{m-1}^{-1}.$

math.DS↗

arXiv2020

Design Principles Developed through User-Centered and Socio-Technical Methods Improve Clinician Satisfaction, Speed, and Confidence in Pharmacogenomic Clinical Decision Support

OBJECTIVE: To design and evaluate new pharmacogenomic (PGx) clinical decision support (CDS) alerts, built to adhere to PGx CDS design principles developed through socio-technical approaches. MATERIALS AND METHODS: Based on previously identified design principles, we created 11 new PGx CDS alert designs and developed an interactive web application containing realistic clinical scenarios and user workflows that mimicked a real-world EHR system. We recruited General Internal Medicine and Cardiology clinicians from Northwestern Medicine and recorded their interactions with the original and new designs. We measured clinician response, satisfaction, speed, and confidence through questionnaires and analysis of the recordings. RESULTS: The study included 12 clinicians. Participants were significantly more satisfied (p=0.0000001), faster (p=0.009), and more confident (p<.05) with the new designs than the original ones. The study lacked statistical power to determine whether prescribing accuracy was improved, but participants were no less accurate, and clinical actions were more concordant with alert interactions (p=0.004) with the new designs. We found a significant learning curve associated with the original designs, which was eliminated with the new designs. DISCUSSION: This study successfully demonstrates that socio-technical and user-centered design techniques can improve PGx CDS alert designs. Best practices for PGx CDS design are limited in the literature, with few effectiveness studies available. These results can help guide future PGx CDS implementations to be more clinician friendly and less time-consuming. CONCLUSION: The results of this study support the PGx CDS design principles we proposed in previous work. As a next step, the new designs should be implemented in a live setting for further validation.

cs.HC↗