اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدUnsupervised Visual Attention and Invariance for Reinforcement Learning
94
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Vision-based reinforcement learning (RL) is successful, but how to generalize it to unknown test environments remains challenging. Existing methods focus on training an RL policy that is universal to changing visual domains, whereas we focus on extracting visual foreground that is universal, feeding clean invariant vision to the RL policy learner. Our method is completely unsupervised, without manual annotations or access to environment internals. Given videos of actions in a training environment, we learn how to extract foregrounds with unsupervised keypoint detection, followed by unsupervised visual attention to automatically generate a foreground mask per video frame. We can then introduce artificial distractors and train a model to reconstruct the clean foreground mask from noisy observations. Only this learned model is needed during test to provide distraction-free visual input to the RL policy learner. Our Visual Attention and Invariance (VAI) method significantly outperforms the state-of-the-art on visual domain generalization, gaining 15 to 49% (61 to 229%) more cumulative rewards per episode on DeepMind Control (our DrawerWorld Manipulation) benchmarks. Our results demonstrate that it is not only possible to learn domain-invariant vision without any supervision, but freeing RL from visual distractions also makes the policy more focused and thus far better.
قيم البحث
اقرأ أيضاً
While an exciting diversity of new imaging devices is emerging that could dramatically improve robotic perception, the challenges of calibrating and interpreting these cameras have limited their uptake in the robotics community. In this work we gener
alise techniques from unsupervised learning to allow a robot to autonomously interpret new kinds of cameras. We consider emerging sparse light field (LF) cameras, which capture a subset of the 4D LF function describing the set of light rays passing through a plane. We introduce a generalised encoding of sparse LFs that allows unsupervised learning of odometry and depth. We demonstrate the proposed approach outperforming monocular and conventional techniques for dealing with 4D imagery, yielding more accurate odometry and depth maps and delivering these with metric scale. We anticipate our technique to generalise to a broad class of LF and sparse LF cameras, and to enable unsupervised recalibration for coping with shifts in camera behaviour over the lifetime of a robot. This work represents a first step toward streamlining the integration of new kinds of imaging devices in robotics applications.
We present an active visual search model for finding objects in unknown environments. The proposed algorithm guides the robot towards the sought object using the relevant stimuli provided by the visual sensors. Existing search strategies are either p
urely reactive or use simplified sensor models that do not exploit all the visual information available. In this paper, we propose a new model that actively extracts visual information via visual attention techniques and, in conjunction with a non-myopic decision-making algorithm, leads the robot to search more relevant areas of the environment. The attention module couples both top-down and bottom-up attention models enabling the robot to search regions with higher importance first. The proposed algorithm is evaluated on a mobile robot platform in a 3D simulated environment. The results indicate that the use of visual attention significantly improves search, but the degree of improvement depends on the nature of the task and the complexity of the environment. In our experiments, we found that performance enhancements of up to 42% in structured and 38% in highly unstructured cluttered environments can be achieved using visual attention mechanisms.
Learning sensorimotor control policies from high-dimensional images crucially relies on the quality of the underlying visual representations. Prior works show that structured latent space such as visual keypoints often outperforms unstructured repres
entations for robotic control. However, most of these representations, whether structured or unstructured are learned in a 2D space even though the control tasks are usually performed in a 3D environment. In this work, we propose a framework to learn such a 3D geometric structure directly from images in an end-to-end unsupervised manner. The input images are embedded into latent 3D keypoints via a differentiable encoder which is trained to optimize both a multi-view consistency loss and downstream task objective. These discovered 3D keypoints tend to meaningfully capture robot joints as well as object movements in a consistent manner across both time and 3D space. The proposed approach outperforms prior state-of-art methods across a variety of reinforcement learning benchmarks. Code and videos at https://buoyancy99.github.io/unsup-3d-keypoints/
In this paper, we propose an end-to-end self-driving network featuring a sparse attention module that learns to automatically attend to important regions of the input. The attention module specifically targets motion planning, whereas prior literatur
e only applied attention in perception tasks. Learning an attention mask directly targeted for motion planning significantly improves the planner safety by performing more focused computation. Furthermore, visualizing the attention improves interpretability of end-to-end self-driving.
For an autonomous agent to fulfill a wide range of user-specified goals at test time, it must be able to learn broadly applicable and general-purpose skill repertoires. Furthermore, to provide the requisite level of generality, these skills must hand
le raw sensory input such as images. In this paper, we propose an algorithm that acquires such general-purpose skills by combining unsupervised representation learning and reinforcement learning of goal-conditioned policies. Since the particular goals that might be required at test-time are not known in advance, the agent performs a self-supervised practice phase where it imagines goals and attempts to achieve them. We learn a visual representation with three distinct purposes: sampling goals for self-supervised practice, providing a structured transformation of raw sensory inputs, and computing a reward signal for goal reaching. We also propose a retroactive goal relabeling scheme to further improve the sample-efficiency of our method. Our off-policy algorithm is efficient enough to learn policies that operate on raw image observations and goals for a real-world robotic system, and substantially outperforms prior techniques.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
