arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2203.11024cs.AIcs.ROcs.SYeess.SY
收藏

Multi-View Dreaming:基于对比学习的多视角世界模型

Multi-View Dreaming: Multi-View World Model with Contrastive Learning

Panasonic Corporation(松下公司);Ritsumeikan University(立命馆大学)

Akira Kinose, Masashi Okada, Ryo Okumura, Tadahiro Taniguchi

摘要:

在本文中,我们提出了多视图的梦想,一种新的强化学习代理集成识别和控制从多个视图的意见,通过扩展梦想。当前的大多数强化学习方法假设单视图观察空间,这对观察到的数据施加了限制,例如缺乏空间信息和遮挡。这使得从环境中获得理想的观测信息变得困难,并且是现实世界机器人应用的瓶颈。在本文中,我们使用对比学习训练共享的潜在空间之间的不同观点,并显示如何产品的专家方法可以用来集成和控制的概率分布的潜在状态的多个观点。我们还提出了Multi-View DreamingV 2,这是Multi-View Dreaming的一个变体,它使用分类分布来模拟潜在状态,而不是高斯分布。实验表明,该方法优于现有的方法在一个现实的机器人控制任务的简单扩展。

英文摘要:

In this paper, we propose Multi-View Dreaming, a novel reinforcement learning agent for integrated recognition and control from multi-view observations by extending Dreaming. Most current reinforcement learning method assumes a single-view observation space, and this imposes limitations on the observed data, such as lack of spatial information and occlusions. This makes obtaining ideal observational information from the environment difficult and is a bottleneck for real-world robotics applications. In this paper, we use contrastive learning to train a shared latent space between different viewpoints, and show how the Products of Experts approach can be used to integrate and control the probability distributions of latent states for multiple viewpoints. We also propose Multi-View DreamingV2, a variant of Multi-View Dreaming that uses a categorical distribution to model the latent state instead of the Gaussian distribution. Experiments show that the proposed method outperforms simple extensions of existing methods in a realistic robot control task.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑