Sim2Real View Invariant Visual Servoing by Recurrent Control

By Fereshteh Sadeghi, Alexander Toshev, Eric Jang, Sergey Levine. 2017

102 citations recorded by the source corpus.

Humans are remarkably proficient at controlling their limbs and tools from a wide range of viewpoints and angles, even in the presence of optical distortions. In robotics, this ability is referred to as visual servoing: moving a tool or end-point to a desired location using primarily visual feedback. In this paper, we study how viewpoint-invariant visual servoing skills can be learned automatically in a robotic manipulation scenario. To this end, we train a deep recurrent controller that can automatically determine which actions move the end-point of a robotic arm to a desired object. The problem that must be solved by this controller is fundamentally ambiguous: under severe variation in viewpoint, it may be impossible to determine the actions in a single feedforward operation. Instead, our visual servoing system must use its memory of past movements to understand how the actions affect the robot motion from the current viewpoint, correcting mistakes and gradually moving closer to the target. This ability is in stark contrast to most visual servoing methods, which either assume known dynamics or require a calibration phase. We show how we can learn this recurrent controller using simulated data and a reinforcement learning objective. We then describe how the resulting model can be transferred to a real-world robot by disentangling perception from control and only adapting the visual layers. The adapted model can servo to previously unseen objects from novel viewpoints on a real-world Kuka IIWA robotic arm. For supplementary videos, see: this https URL

Subfields: Biophysics, Computer Vision and Pattern Recognition, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biophysics

In Computer Vision and Pattern Recognition

In Media Technology

Citations in the corpus

Dexterous Manipulation with Deep Reinforcement Learning: Efficient, General, and Low-Cost (2018, methodology): “This approach has been employed for manipulation [24], visual servoing [25], locomotion [26], and navigation [23].”
Learning by Playing - Solving Sparse Reward Tasks from Scratch (2018, background): “These include for example: shaping rewards (Ng et al., 1999; Randløv & Alstrøm, 1998; Gu et al., 2017), curriculum learning (Heess et al., 2017; Ghosh et al., 2018; Forestier et al., 2017), transfer of learned policies from simulation to reality (see Duan et al. (2017); Sadeghi et al. (2017); Tobin…”
BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning (2022, background): “This includes works that generalize skills to novel objects [31, 32, 33, 34, 35], to novel 76 environments [36], from simulation to reality [37, 38, 39, 40, 41], and to new manipulation skills 77 and objects [42, 43, 44, 45].”
SplitNet: Sim2Sim and Task2Task Transfer for Embodied Visual Navigation (2019, background): “[27] randomize colors, textures, lighting, and camera pose as a form of augmentation of the simulated imagery to better generalize to real-world imagery, but focus on primitive geometric objects for picking tasks.”
RetinaGAN: An Object-aware Approach to Sim-to-Real Transfer (2020, methodology): “With domain randomization, a policy is trained with randomized simulation parameters and scene configurations which produce differences in visual appearance [6], [7], [8], [9], [10], [11].”
IKEA Furniture Assembly Environment for Long-Horizon Complex Manipulation Tasks (2019, background): “Once trained with enough variation, the policy should be able to generalize to the real world [30–32, 22, 5].”
Visual Representations for Semantic Target Driven Navigation (2018, background): “et al. [14]) for point goal and room goal prediction are benchmarked in the Minos environment [15]. For the related problems of object reaching with a manipulator and quadcopter ﬂying, Sadeghi et al. [16], [17] use synthetic data with domain randomization to learn control strategies. Gupta et al. [1”
Action Image Representation: Learning Scalable Deep Grasping Policies with Zero Real World Data (2020, methodology): “Researchers have used domain randomization [23], [22], [10] or trained pixel-to-pixel networks, converting either from simulated images to real images [4], or vice versa [11].”