4 February 2019 in Scientific Reports
Predicting physical response of an artificially structured material is of particular interest for scientific and engineering applications. Here we use deep learning to predict optical response of artificially engineered nanophotonic devices. In addition to predicting forward approximation of transmission response for any given topology, this approach allows us to inversely approximate designs for a targeted optical response. Our Deep Neural Network (DNN) could design compact (2.6 × 2.6 μm2) silicon-on-insulator (SOI)-based 1 × 2 power splitters with various target splitting ratios in a fraction of a second. This model is trained to minimize the reflection (to smaller than ~ −20 dB) while achieving maximum transmission efficiency above 90% and target splitting specifications. This approach paves the way for rapid design of integrated photonic components relying on complex nanostructures.
8 January 2021 in Journal of Lightwave Technology
Deep learning is now playing a major role in designing photonic devices, including nanostructured photonics. In this paper, we investigate three models for designing nanophonic power splitters with multiple splitting ratios. The first model is a forward regression model, wherein the trained deep neural network (DNN) is used within the optimization loop. The second is an inverse regression model, in which the trained DNN constructs a structure with the desired target performance given as input. The third model is a generative network, which can randomly produce a series of optimized designs for a target performance. Focusing on the nanophotonic power splitters, we show how the devices can be designed by these three types of DNN models.
5 July 2023 in IEEE Robotics and Automation Letters
GelSight sensors that estimate contact geometry and force by reconstructing the deformation of their soft elastomer from images would yield poor force measurements when the elastomer deforms uniformly or reaches deformation saturation. Here we present an L $^{3}$ F-TOUCH sensor that considerably enhances the three-axis force sensing capability of typical GelSight sensors. Specifically, the L $^{3}$ F-TOUCH sensor comprises: (i) an elastomer structure resembling the classic GelSight sensor design for fine-grained contact geometry sensing; and (ii) a mechanically simple suspension structure to enable three-dimensional elastic displacement of the elastomer structure upon contact. Such displacement is tracked by detecting the displacement of an ARTag and is transformed to three-axis contact force via calibration. We further revamp the sensor's optical system by fixing the ARTag on the base and reflecting it to the same camera viewing the elastomer through a mirror. As a result, the tactile and force sensing modes can operate independently, but the entire L $^{3}$ F-TOUCH remains L ight-weight and L ow-cost while facilitating a wire L ess deployment. Evaluations and experiment results demonstrate that the proposed L $^{3}$ F-TOUCH sensor compromises GelSight's limitation in force sensing and is more practical compared with equipping commercial three-axis force sensors. Thus, the L $^{3}$ F-TOUCH could further empower existing Vision-based Tactile Sensors (VBTSs) in replication and deployment.
5 June 2024 in Machine Learning
The success of deep learning in computer vision and natural language processing communities can be attributed to the training of very deep neural networks with millions or billions of parameters, which can then be trained with massive amounts of data. However, a similar trend has largely eluded the training of deep reinforcement learning (RL) algorithms where larger networks do not lead to performance improvement. Previous work has shown that this is mostly due to instability during the training of deep RL agents when using larger networks. In this paper, we make an attempt to understand and address the training of larger networks for deep RL. We first show that naively increasing network capacity does not improve performance. Then, we propose a novel method that consists of (1) wider networks with DenseNet connection, (2) decoupling representation learning from the training of RL, and (3) a distributed training method to mitigate overfitting problems. Using this three-fold technique, we show that we can train very large networks that result in significant performance gains. We present several ablation studies to demonstrate the efficacy of the proposed method and some intuitive understanding of the reasons for performance gain. We show that our proposed method outperforms other baseline algorithms on several challenging locomotion tasks.
28 February 2020 in IEEE Robotics and Automation Letters
In this paper, we propose a derivative-free model learning framework for Reinforcement Learning (RL) algorithms based on Gaussian Process Regression (GPR). In many mechanical systems, only positions can be measured by the sensing instruments. Then, instead of representing the system state as suggested by the physics with a collection of positions, velocities, and accelerations, we define the state as the set of past position measurements. However, the equation of motions derived by physical first principles cannot be directly applied in this framework, being functions of velocities and accelerations. For this reason, we introduce a novel derivative-free physically-inspired kernel, which can be easily combined with nonparametric derivative-free Gaussian Process models. Tests performed on two real platforms show that the considered state definition combined with the proposed model improves estimation performance and data-efficiency w.r.t. traditional models based on GPR. Finally, we validate the proposed framework by solving two RL control problems for two real robotic systems.