Search

In this paper, we present a Stochastic Deep Neural Network-based Model Reference Adaptive Control. Building on our work "Deep Model Reference Adaptive Control", we extend the controller capability by using Bayesian deep neural networks (DNN) to represent uncertainties and model non-linearities. Stochastic Deep Model Reference Adaptive Control uses a Lyapunov-based method to adapt the output-layer weights of the DNN model in real-time, while a data-driven supervised learning algorithm is used to update the inner-layers parameters. This asynchronous network update ensures boundedness and guaranteed tracking performance with a learning-based real-time feedback controller. A Bayesian approach to DNN learning helped avoid over-fitting the data and provide confidence intervals over the predictions. The controller's stochastic nature also ensured "Induced Persistency of excitation," leading to convergence of the overall system signal., Comment: Accepted in IEEE CDC 2021

Efficient and robust policy transfer remains a key challenge for reinforcement learning to become viable for real-wold robotics. Policy transfer through warm initialization, imitation, or interacting over a large set of agents with randomized instances, have been commonly applied to solve a variety of Reinforcement Learning tasks. However, this seems far from how skill transfer happens in the biological world: Humans and animals are able to quickly adapt the learned behaviors between similar tasks and learn new skills when presented with new situations. Here we seek to answer the question: Will learning to combine adaptation and exploration lead to a more efficient transfer of policies between domains? We introduce a principled mechanism that can "Adapt-to-Learn", that is adapt the source policy to learn to solve a target task with significant transition differences and uncertainties. We show that the presented method learns to seamlessly combine learning from adaptation and exploration and leads to a robust policy transfer algorithm with significantly reduced sample complexity in transferring skills between related tasks.

In this paper, we present Asynchronous implementation of Deep Neural Network-based Model Reference Adaptive Control (DMRAC). We evaluate this new neuro-adaptive control architecture through flight tests on a small quadcopter. We demonstrate that a single DMRAC controller can handle significant nonlinearities due to severe system faults and deliberate wind disturbances while executing high-bandwidth attitude control. We also show that the architecture has long-term learning abilities across different flight regimes, and can generalize to fly different flight trajectories than those on which it was trained. These results demonstrating the efficacy of this architecture for high bandwidth closed-loop attitude control of unstable and nonlinear robots operating in adverse situations. To achieve these results, we designed a software+communication architecture to ensure online real-time inference of the deep network on a high-bandwidth computation-limited platform. We expect that this architecture will benefit other deep learning in the closed-loop experiments on robots., Comment: Accepted in CORL2020. arXiv admin note: text overlap with arXiv:1909.08602

The main focus of the work presented in this thesis is to develop an optimal control based formation flying control strategy for high precision formation flying of small satellites that have restricted computation and storage capacity. Using the recently developed model predictive static programming (MPSP), and Generalized MPSP algorithm a suboptimal guidance logic is presented for the formation flying of small satellites. The proposed guidance scheme is valid both for high eccentricity chief satellite orbits as well as the large separation distance between the chief and deputy satellites. Comparative study with standard Linear Quadratic Regulator (LQR) solution (which serves as a guess solution for MPSP) and another nonlinear controller, Finite-time State-Dependent Ricatti Equation (SDRE) reveals that MPSP guidance achieves the objective with higher accuracy and with a lesser amount of control usage.

We present a new neuroadaptive architecture: Deep Neural Network based Model Reference Adaptive Control (DMRAC). Our architecture utilizes the power of deep neural network representations for modeling significant nonlinearities while marrying it with the boundedness guarantees that characterize MRAC based controllers. We demonstrate through simulations and analysis that DMRAC can subsume previously studied learning based MRAC methods, such as concurrent learning and GP-MRAC. This makes DMRAC a highly powerful architecture for high-performance control of nonlinear systems with long-term learning properties., Comment: Accepted in IEEE CDC-2019

In this paper, we present a hybrid direct-indirect model reference adaptive controller (MRAC), to address a class of problems with matched and unmatched uncertainties. In the proposed architecture, the unmatched uncertainty is estimated online through a companion observer model. Upon convergence of the observer, the unmatched uncertainty estimate is remodeled into a state dependent linear form to augment the nominal system dynamics. Meanwhile, a direct adaptive controller designed for a switching system cancels the effect of matched uncertainty in the system and achieves reference model tracking. We demonstrate that the proposed hybrid controller can handle a broad class of nonlinear systems with both matched and unmatched uncertainties, Comment: Accepted at CODIT 2019

In this paper, we present a new approach to Transfer Learning (TL) in Reinforcement Learning (RL) for cross-domain tasks. Many of the available techniques approach the transfer architecture as a method of speeding up the target task learning. We propose to adapt and reuse the mapped source task optimal-policy directly in related domains. We show the optimal policy from a related source task can be near optimal in target domain provided an adaptive policy accounts for the model error between target and source. The main benefit of this policy augmentation is generalizing policies across multiple related domains without having to re-learn the new tasks. Our results show that this architecture leads to better sample efficiency in the transfer, reducing sample complexity of target task learning to target apprentice learning., Comment: To appear as conference paper in ICRA 2018

Atomic Force Microscopy (AFM) and Ultrasonic Force Microscopy (UFM) have been applied to the characterization of composite samples formed by SrTiO 3 (STO) nanoparticles (NPs) and polyvinyl alcohol (PVA). The morphological features of the STO NPs were much better resolved in UFM than in contact-mode AFM topography. For high STO concentrations the individual STO NPs formed nanoclusters, which gathered in microaggregates. The STO aggregates, covered by PVA, exhibited no AFM frictional contrast, but were clearly distinguished from the PVA matrix using UFM. Similar aggregation was observed for NPs in the composite samples than for NPs deposited on top of a flat silicon substrate from a milliQ water solution in the absence of polymer. In the hybrid films, most STO nanoparticles typically presented a lower UFM contrast than the PVA matrix, even though stiffer sample regions such as STO should give rise to a higher UFM contrast. STO NPs with intermediate contrast were characterized by an UFM halo of lower contrast at the PVA/STO interface. The results may be explained by considering that ultrasound is effectively damped on the nanometer scale at PVA/ STO interfaces. According to our data, the nanoscale ultrasonic response at the PVA/STO interface plays a fundamental role in the UFM image contrast.