Showing 1–36 of 36 results for author: Kundu, A

RepCNN: Micro-sized, Mighty Models for Wakeword Detection

Authors: Arnav Kundu, Prateeth Nayak, Hywel Richards, Priyanka Padmanabhan, Devang Naik

Abstract: Always-on machine learning models require a very low memory and compute footprint. Their restricted parameter count limits the model's capacity to learn, and the effectiveness of the usual training algorithms to find the best parameters. Here we show that a small convolutional model can be better trained by first refactoring its computation into a larger redundant multi-branched architecture. Then… ▽ More Always-on machine learning models require a very low memory and compute footprint. Their restricted parameter count limits the model's capacity to learn, and the effectiveness of the usual training algorithms to find the best parameters. Here we show that a small convolutional model can be better trained by first refactoring its computation into a larger redundant multi-branched architecture. Then, for inference, we algebraically re-parameterize the trained model into the single-branched form with fewer parameters for a lower memory footprint and compute cost. Using this technique, we show that our always-on wake-word detector model, RepCNN, provides a good trade-off between latency and accuracy during inference. RepCNN re-parameterized models are 43% more accurate than a uni-branch convolutional model while having the same runtime. RepCNN also meets the accuracy of complex architectures like BC-ResNet, while having 2x lesser peak memory usage and 10x faster runtime. △ Less

Submitted 4 June, 2024; originally announced June 2024.

Stabilizing switched nonlinear systems under restricted but arbitrary switching signals

Authors: Atreyee Kundu

Abstract: This paper deals with input/output-to-state stability (IOSS) of switched nonlinear systems whose switching signals obey pre-specified restrictions on admissible switches between the subsystems and admissible dwell times on the subsystems. We present sufficient conditions on the subsystems, admissible switches between them and admissible dwell times on them, such that a switched system generated un… ▽ More This paper deals with input/output-to-state stability (IOSS) of switched nonlinear systems whose switching signals obey pre-specified restrictions on admissible switches between the subsystems and admissible dwell times on the subsystems. We present sufficient conditions on the subsystems, admissible switches between them and admissible dwell times on them, such that a switched system generated under all switching signals obeying the given restrictions is IOSS. Multiple Lyapunov-like functions and graph theory are the key apparatuses for our analysis. A numerical example is presented to demonstrate our results. △ Less

Submitted 3 April, 2024; originally announced April 2024.

Comments: Work under review. arXiv admin note: text overlap with arXiv:2306.10564

A Sparsity Approach to Scheduling and Control of Networked Systems

Authors: Anubhab Dasgupta, Atreyee Kundu

Abstract: We study the design of scheduling logic and control logic for networked control systems (NCSs) where plants communicate with their remotely located controllers over a shared band-limited communication network. Due to a limited capacity of the network, only a subset of the plants can exchange information with their controllers at any instant of time and the remaining plants operate in open-loop. Ou… ▽ More We study the design of scheduling logic and control logic for networked control systems (NCSs) where plants communicate with their remotely located controllers over a shared band-limited communication network. Due to a limited capacity of the network, only a subset of the plants can exchange information with their controllers at any instant of time and the remaining plants operate in open-loop. Our key contribution is a new algorithm that co-designs (a) an allocation scheme of the communication network among the plants (scheduling logic) and (b) the control inputs for the plants accessing the network (control logic) under which given non-zero initial states are steered to zero in a given time horizon for all the plants in the NCS. Sparse optimization is the primary apparatus for our analysis. We also provide sufficient conditions on the plant dynamics, capacity of the communication network and the given time horizon that lead to a numerically tractable implementation of our algorithm. A numerical experiment is presented to demonstrate the proposed results. △ Less

Submitted 5 December, 2023; originally announced December 2023.

Comments: 19 pages, 3 figures

A probabilistic algorithm for scheduling networked control systems under data losses

Authors: Anubhab Dasgupta, Darsana Udayakumar, Atreyee Kundu

Abstract: This paper deals with the design of scheduling logics for networked control systems (NCSs) whose communication networks have limited capacity and are prone to data losses. Our contributions are twofold. First, we present a probabilistic algorithm to generate a scheduling logic that under certain conditions on the plant and the controller dynamics, the capacity of the network and the probability of… ▽ More This paper deals with the design of scheduling logics for networked control systems (NCSs) whose communication networks have limited capacity and are prone to data losses. Our contributions are twofold. First, we present a probabilistic algorithm to generate a scheduling logic that under certain conditions on the plant and the controller dynamics, the capacity of the network and the probability of data losses, ensures stochastic stability of each plant in the NCS. Second, given the plant dynamics, the capacity of the shared communication network and the probability of data losses, we discuss the design of state-feedback controllers such that our stability conditions are obeyed. Numerical examples are presented to demonstrate the results reported in this paper. △ Less

Submitted 27 June, 2023; originally announced June 2023.

Comments: 8 pages, 5 figures. arXiv admin note: text overlap with arXiv:2204.00788

Journal ref: Indian Control Conference 2023

On stability and state-norm estimation of switched systems under restricted switching

Authors: Atreyee Kundu

Abstract: This paper deals with the analysis of input/output-to-state stability (IOSS) and construction of state-norm estimators for continuous-time switched nonlinear systems under restricted switching. Our contributions are twofold. First, given a family of systems, possibly containing unstable dynamics, a set of admissible switches between the subsystems and admissible minimum and maximum dwell times on… ▽ More This paper deals with the analysis of input/output-to-state stability (IOSS) and construction of state-norm estimators for continuous-time switched nonlinear systems under restricted switching. Our contributions are twofold. First, given a family of systems, possibly containing unstable dynamics, a set of admissible switches between the subsystems and admissible minimum and maximum dwell times on the subsystems, we identify a class of switching signals that obeys the given restrictions and preserves IOSS of the resulting switched system. Second, we design a class of state-norm estimators for switched systems under our class of stabilizing switching signals. These estimators are switched systems themselves with two subsystems -- one stable and one unstable. The key apparatus for our analysis is multiple Lyapunov-like functions. A numerical example is presented to demonstrate the results. △ Less

Submitted 23 April, 2024; v1 submitted 18 June, 2023; originally announced June 2023.

Comments: 17 pages, 4 figures. Longer version of a manuscript under review. arXiv admin note: text overlap with arXiv:2207.07764

R2 Loss: Range Restriction Loss for Model Compression and Quantization

Authors: Arnav Kundu, Chungkuk Yoo, Srijan Mishra, Minsik Cho, Saurabh Adya

Abstract: Model quantization and compression is widely used techniques to reduce usage of computing resource at inference time. While state-of-the-art works have been achieved reasonable accuracy with higher bit such as 4bit or 8bit, but still it is challenging to quantize/compress a model further, e.g., 1bit or 2bit. To overcome the challenge, we focus on outliers in weights of a pre-trained model which di… ▽ More Model quantization and compression is widely used techniques to reduce usage of computing resource at inference time. While state-of-the-art works have been achieved reasonable accuracy with higher bit such as 4bit or 8bit, but still it is challenging to quantize/compress a model further, e.g., 1bit or 2bit. To overcome the challenge, we focus on outliers in weights of a pre-trained model which disrupt effective lower bit quantization and compression. In this work, we propose Range Restriction Loss (R2-Loss) for building lower bit quantization and compression friendly models by removing outliers from weights during pre-training. By effectively restricting range of weights, we mold the overall distribution into a tight shape to ensure high quantization bit resolution, therefore allowing model compression and quantization techniques can to utilize their limited numeric representation powers better. We introduce three different, L-inf R2-Loss, its extension Margin R2-Loss and a new Soft-Min-MaxR2-Loss to be used as an auxiliary loss during full-precision model training. These R2-Loss can be used in different cases such as L-inf and Margin R2-Loss would be effective for symmetric quantization, while Soft-Min-Max R2-Loss shows better performance for model compression. In our experiment, R2-Loss improves lower bit quantization accuracy with state-of-the-art post-training quantization (PTQ), quantization-aware training (QAT), and model compression techniques. With R2-Loss, MobileNet-V2 2bit weight and 8bit activation PTQ, MobileNet-V1 2bit weight and activation QAT, ResNet18 1bit weight compression are improved to 59.49% from 50.66%, 59.05% from 55.96%, and 52.58% from 45.54%, respectively. △ Less

Submitted 11 February, 2024; v1 submitted 14 March, 2023; originally announced March 2023.

Analysis of Arrhythmia Classification on ECG Dataset

Authors: Taminul Islam, Arindom Kundu, Tanzim Ahmed, Nazmul Islam Khan

Abstract: The heart is one of the most vital organs in the human body. It supplies blood and nutrients in other parts of the body. Therefore, maintaining a healthy heart is essential. As a heart disorder, arrhythmia is a condition in which the heart's pum** mechanism becomes aberrant. The Electrocardiogram is used to analyze the arrhythmia problem from the ECG signals because of its fewer difficulties and… ▽ More The heart is one of the most vital organs in the human body. It supplies blood and nutrients in other parts of the body. Therefore, maintaining a healthy heart is essential. As a heart disorder, arrhythmia is a condition in which the heart's pum** mechanism becomes aberrant. The Electrocardiogram is used to analyze the arrhythmia problem from the ECG signals because of its fewer difficulties and cheapness. The heart peaks shown in the ECG graph are used to detect heart diseases, and the R peak is used to analyze arrhythmia disease. Arrhythmia is grouped into two groups - Tachycardia and Bradycardia for detection. In this paper, we discussed many different techniques such as Deep CNNs, LSTM, SVM, NN classifier, Wavelet, TQWT, etc., that have been used for detecting arrhythmia using various datasets throughout the previous decade. This work shows the analysis of some arrhythmia classification on the ECG dataset. Here, Data preprocessing, feature extraction, classification processes were applied on most research work and achieved better performance for classifying ECG signals to detect arrhythmia. Automatic arrhythmia detection can help cardiologists make the right decisions immediately to save human life. In addition, this research presents various previous research limitations with some challenges in detecting arrhythmia that will help in future research. △ Less

Submitted 10 January, 2023; originally announced January 2023.

Comments: 6 pages, 5 figures. This paper has been published to 2022 proceedings of IEEE 7th International conference for Convergence in Technology (I2CT), 07-09 April 2022, Mumbai, India

Journal ref: In 2022 IEEE 7th International conference for Convergence in Technology (I2CT) (pp. 1-6). IEEE

HEiMDaL: Highly Efficient Method for Detection and Localization of wake-words

Authors: Arnav Kundu, Mohammad Samragh Razlighi, Minsik Cho, Priyanka Padmanabhan, Devang Naik

Abstract: Streaming keyword spotting is a widely used solution for activating voice assistants. Deep Neural Networks with Hidden Markov Model (DNN-HMM) based methods have proven to be efficient and widely adopted in this space, primarily because of the ability to detect and identify the start and end of the wake-up word at low compute cost. However, such hybrid systems suffer from loss metric mismatch when… ▽ More Streaming keyword spotting is a widely used solution for activating voice assistants. Deep Neural Networks with Hidden Markov Model (DNN-HMM) based methods have proven to be efficient and widely adopted in this space, primarily because of the ability to detect and identify the start and end of the wake-up word at low compute cost. However, such hybrid systems suffer from loss metric mismatch when the DNN and HMM are trained independently. Sequence discriminative training cannot fully mitigate the loss-metric mismatch due to the inherent Markovian style of the operation. We propose an low footprint CNN model, called HEiMDaL, to detect and localize keywords in streaming conditions. We introduce an alignment-based classification loss to detect the occurrence of the keyword along with an offset loss to predict the start of the keyword. HEiMDaL shows 73% reduction in detection metrics along with equivalent localization accuracy and with the same memory footprint as existing DNN-HMM style models for a given wake-word. △ Less

Submitted 26 October, 2022; originally announced October 2022.

I see what you hear: a vision-inspired method to localize words

Authors: Mohammad Samragh, Arnav Kundu, Ting-Yao Hu, Minsik Cho, Aman Chadha, Ashish Shrivastava, Oncel Tuzel, Devang Naik

Abstract: This paper explores the possibility of using visual object detection techniques for word localization in speech data. Object detection has been thoroughly studied in the contemporary literature for visual data. Noting that an audio can be interpreted as a 1-dimensional image, object localization techniques can be fundamentally useful for word localization. Building upon this idea, we propose a lig… ▽ More This paper explores the possibility of using visual object detection techniques for word localization in speech data. Object detection has been thoroughly studied in the contemporary literature for visual data. Noting that an audio can be interpreted as a 1-dimensional image, object localization techniques can be fundamentally useful for word localization. Building upon this idea, we propose a lightweight solution for word detection and localization. We use bounding box regression for word localization, which enables our model to detect the occurrence, offset, and duration of keywords in a given audio stream. We experiment with LibriSpeech and train a model to localize 1000 words. Compared to existing work, our method reduces model size by 94%, and improves the F1 score by 6.5\%. △ Less

Submitted 24 October, 2022; originally announced October 2022.

Yet another stability condition for switched nonlinear systems

Authors: Atreyee Kundu

Abstract: This paper deals with input/output-to-state stability (IOSS) of continuous-time switched nonlinear systems. Given a family of systems, possibly containing unstable dynamics, and a set of restrictions on admissible switches between the subsystems and admissible dwell times on the subsystems, we identify a class of switching signals that obeys these restrictions and preserves stability of the result… ▽ More This paper deals with input/output-to-state stability (IOSS) of continuous-time switched nonlinear systems. Given a family of systems, possibly containing unstable dynamics, and a set of restrictions on admissible switches between the subsystems and admissible dwell times on the subsystems, we identify a class of switching signals that obeys these restrictions and preserves stability of the resulting switched system. The primary apparatus for our analysis is multiple Lyapunov-like functions. Input-to-state stability (ISS) and global asymptotic stability (GAS) of switched systems under pre-specified restrictions on switching signals fall as special cases of our results when no outputs (resp., also inputs) are considered. △ Less

Submitted 16 June, 2023; v1 submitted 15 July, 2022; originally announced July 2022.

Comments: 10 pages, 1 figure

A probabilistic scheduling algorithm for networked control systems

Authors: Meghna Singh, Atreyee Kundu

Abstract: This paper deals with the design of scheduling logics for Networked Control Systems (NCSs) whose communication networks have limited capacity. We assume that only a subset of the plants can communicate with their controllers at any time instant. Our contributions are twofold. First, we present a probabilistic algorithm to design scheduling logics that, under certain conditions on the plant and con… ▽ More This paper deals with the design of scheduling logics for Networked Control Systems (NCSs) whose communication networks have limited capacity. We assume that only a subset of the plants can communicate with their controllers at any time instant. Our contributions are twofold. First, we present a probabilistic algorithm to design scheduling logics that, under certain conditions on the plant and controller dynamics and the capacity of the network, ensure stochastic stability of each plant in an NCS. Second, given the plant dynamics and the capacity of the network, we design static state-feedback controllers such that the conditions for stability under our scheduling logics are satisfied. The main apparatus for our analysis is a Markovian jump linear system representation of the individual plants in an NCS. Our stability conditions involve sets of matrix inequalities. We present numerical experiments to demonstrate our results. △ Less

Submitted 2 April, 2022; originally announced April 2022.

Comments: A slightly longer version of our manuscript under review

Journal ref: IFAC Conference on Networked Systems 2022

Quadratic and Higher-Order Unconstrained Binary Optimization of Railway Rescheduling for Quantum Computing

Authors: Krzysztof Domino, Akash Kundu, Özlem Salehi, Krzysztof Krawiec

Abstract: As consequences of disruptions in railway traffic affect passenger experience/satisfaction, appropriate rerouting and/or rescheduling is necessary. These problems are known to be NP-hard, given the numerous restrictions of traffic nature. With the recent advances in quantum technologies, quantum annealing has become an alternative method to solve such optimization problems. To use quantum annealin… ▽ More As consequences of disruptions in railway traffic affect passenger experience/satisfaction, appropriate rerouting and/or rescheduling is necessary. These problems are known to be NP-hard, given the numerous restrictions of traffic nature. With the recent advances in quantum technologies, quantum annealing has become an alternative method to solve such optimization problems. To use quantum annealing, the problem needs to be encoded in QUBO (quadratic unconstrained binary optimization) or HOBO (higher-order binary optimization) formulation that can be recast as a QUBO. This paper introduces QUBO and HOBO representations for rescheduling problems of railway traffic management; the latter is a new approach up to our knowledge. This new approach takes into account not only the single-track lines but also the double- and multi-track lines, as well as stations composed of tracks and switches. We consider the conditions of minimal headway between trains, minimal stay on stations, track occupation, and rolling stock circulation. Furthermore, a hybrid quantum-classical procedure is presented that includes rerouting. We demonstrate the proof of concept implementation on the D-Wave Quantum Processing Unit and D-Wave hybrid solver. △ Less

Submitted 20 July, 2022; v1 submitted 7 July, 2021; originally announced July 2021.

Journal ref: Quantum Information Processing volume 21, Article number: 337 (2022)

A scheduling algorithm for networked control systems

Authors: Atreyee Kundu

Abstract: This paper deals with the design of scheduling logics for Networked Control Systems (NCSs) whose shared communication networks have limited capacity. We assume that among \(N\) plants, only \(M\:(< N)\) plants can communicate with their controllers at any time instant. We present an algorithm to allocate the network to the plants periodically such that stability of each plant is preserved. The mai… ▽ More This paper deals with the design of scheduling logics for Networked Control Systems (NCSs) whose shared communication networks have limited capacity. We assume that among \(N\) plants, only \(M\:(< N)\) plants can communicate with their controllers at any time instant. We present an algorithm to allocate the network to the plants periodically such that stability of each plant is preserved. The main apparatus for our analysis is a switched systems representation of the individual plants in an NCS. We rely on multiple Lyapunov-like functions and graph-theoretic arguments to design our scheduling logics. The set of results presented in this paper is a continuous-time counterpart of the results proposed in [15]. We present a set of numerical experiments to demonstrate the performance of our techniques. △ Less

Submitted 3 January, 2021; originally announced January 2021.

Comments: 15 pages, 3 figures. A continuous-time counterpart of our earlier result on scheduling of networked control systems. arXiv admin note: substantial text overlap with arXiv:1901.08353, arXiv:2010.08447

Journal ref: IEEE Conference on Decision and Control, 2021

A note on the existence of stabilizing switching signals for switched linear systems

Authors: Atreyee Kundu

Abstract: This paper deals with stability of discrete-time switched linear systems whose all subsystems are unstable. We present sufficient conditions on the subsystems matrices such that a switched system is globally exponentially stable under a set of purely time-dependent switching signals that are allowed to activate all subsystems. The main apparatuses for our analysis are (matrix) commutation relation… ▽ More This paper deals with stability of discrete-time switched linear systems whose all subsystems are unstable. We present sufficient conditions on the subsystems matrices such that a switched system is globally exponentially stable under a set of purely time-dependent switching signals that are allowed to activate all subsystems. The main apparatuses for our analysis are (matrix) commutation relations between certain products of the subsystems matrices and graph-theoretic arguments. We present a numerical experiment to demonstrate our results. △ Less

Submitted 31 October, 2020; originally announced November 2020.

Comments: 11 pages, 2 figures. arXiv admin note: substantial text overlap with arXiv:2005.10493

Journal ref: IFAC Workshop on Analysis and Design of Hybrid Systems, 2021

Optimize what matters: Training DNN-HMM Keyword Spotting Model Using End Metric

Authors: Ashish Shrivastava, Arnav Kundu, Chandra Dhir, Devang Naik, Oncel Tuzel

Abstract: Deep Neural Network--Hidden Markov Model (DNN-HMM) based methods have been successfully used for many always-on keyword spotting algorithms that detect a wake word to trigger a device. The DNN predicts the state probabilities of a given speech frame, while HMM decoder combines the DNN predictions of multiple speech frames to compute the keyword detection score. The DNN, in prior methods, is traine… ▽ More Deep Neural Network--Hidden Markov Model (DNN-HMM) based methods have been successfully used for many always-on keyword spotting algorithms that detect a wake word to trigger a device. The DNN predicts the state probabilities of a given speech frame, while HMM decoder combines the DNN predictions of multiple speech frames to compute the keyword detection score. The DNN, in prior methods, is trained independent of the HMM parameters to minimize the cross-entropy loss between the predicted and the ground-truth state probabilities. The mis-match between the DNN training loss (cross-entropy) and the end metric (detection score) is the main source of sub-optimal performance for the keyword spotting task. We address this loss-metric mismatch with a novel end-to-end training strategy that learns the DNN parameters by optimizing for the detection score. To this end, we make the HMM decoder (dynamic programming) differentiable and back-propagate through it to maximize the score for the keyword and minimize the scores for non-keyword speech segments. Our method does not require any change in the model architecture or the inference framework; therefore, there is no overhead in run-time memory or compute requirements. Moreover, we show significant reduction in false rejection rate (FRR) at the same false trigger experience (> 70% over independent DNN training). △ Less

Submitted 25 February, 2021; v1 submitted 2 November, 2020; originally announced November 2020.

Comments: Accepted at ICASSP 2021

Design of periodic scheduling and control for networked systems under random data loss

Authors: Atreyee Kundu, Daniel E. Quevedo

Abstract: This paper deals with Networked Control Systems (NCSs) whose shared networks have limited communication capacity and are prone to data losses. We assume that among (N) plants, only (M < N) plants can communicate with their controllers at any time instant. In addition, a control input, at any time instant, is lost in a channel with a probability (p). Our contributions are threefold. First, we ident… ▽ More This paper deals with Networked Control Systems (NCSs) whose shared networks have limited communication capacity and are prone to data losses. We assume that among (N) plants, only (M < N) plants can communicate with their controllers at any time instant. In addition, a control input, at any time instant, is lost in a channel with a probability (p). Our contributions are threefold. First, we identify necessary and sufficient conditions on the open-loop and closed-loop dynamics of the plants that ensure existence of purely time-dependent periodic scheduling sequences under which stability of each plant is preserved for all admissible data loss signals. Second, given the open-loop and closed-loop dynamics of the plants, relevant parameters of the shared network and a period for the scheduling sequence, we present an algorithm that verifies our stability conditions and if satisfied, designs stabilizing scheduling sequences. Otherwise, the algorithm reports non-existence of a stabilizing periodic scheduling sequence with the given period and stability margins. Third, given the plant matrices, the parameters of the network and a period for the scheduling sequence, we present an algorithm that designs static state-feedback controllers such that our stability conditions are satisfied. The main apparatus for our analysis is a switched systems representation of the individual plants in an NCS whose switching signals are time-inhomogeneous Markov chains. Our stability conditions rely on the existence of sets of symmetric and positive definite matrices that satisfy certain (in)equalities. △ Less

Submitted 16 October, 2020; originally announced October 2020.

Comments: Extended version of our manuscript under review

Journal ref: IEEE Transactions on Control of Network Systems, early access, 2021

Learning event-driven switched linear systems

Authors: Atreyee Kundu, Pavithra Prabhakar

Abstract: We propose an automata theoretic learning algorithm for the identification of black-box switched linear systems whose switching logics are event-driven. A switched system is expressed by a deterministic finite automaton (FA) whose node labels are the subsystem matrices. With information about the dimensions of the matrices and the set of events, and with access to two oracles, that can simulate th… ▽ More We propose an automata theoretic learning algorithm for the identification of black-box switched linear systems whose switching logics are event-driven. A switched system is expressed by a deterministic finite automaton (FA) whose node labels are the subsystem matrices. With information about the dimensions of the matrices and the set of events, and with access to two oracles, that can simulate the system on a given input, and provide counter-examples when given an incorrect hypothesis automaton, we provide an algorithm that outputs the unknown FA. Our algorithm first uses the oracle to obtain the node labels of the system run on a given input sequence of events, and then extends Angluin's \(L^*\)-algorithm to determine the FA that accepts the language of the given FA. We demonstrate the performance of our learning algorithm on a set of benchmark examples. △ Less

Submitted 27 September, 2020; originally announced September 2020.

Journal ref: Indian Control Conference, 2021

A reinforcement learning approach to hybrid control design

Authors: Meet Gandhi, Atreyee Kundu, Shalabh Bhatnagar

Abstract: In this paper we design hybrid control policies for hybrid systems whose mathematical models are unknown. Our contributions are threefold. First, we propose a framework for modelling the hybrid control design problem as a single Markov Decision Process (MDP). This result facilitates the application of off-the-shelf algorithms from Reinforcement Learning (RL) literature towards designing optimal co… ▽ More In this paper we design hybrid control policies for hybrid systems whose mathematical models are unknown. Our contributions are threefold. First, we propose a framework for modelling the hybrid control design problem as a single Markov Decision Process (MDP). This result facilitates the application of off-the-shelf algorithms from Reinforcement Learning (RL) literature towards designing optimal control policies. Second, we model a set of benchmark examples of hybrid control design problem in the proposed MDP framework. Third, we adapt the recently proposed Proximal Policy Optimisation (PPO) algorithm for the hybrid action space and apply it to the above set of problems. It is observed that in each case the algorithm converges and finds the optimal policy. △ Less

Submitted 2 September, 2020; originally announced September 2020.

Comments: 9 pages

Virtual Multi-view Fusion for 3D Semantic Segmentation

Authors: Abhijit Kundu, Xiaoqi Yin, Alireza Fathi, David Ross, Brian Brewington, Thomas Funkhouser, Caroline Pantofaru

Abstract: Semantic segmentation of 3D meshes is an important problem for 3D scene understanding. In this paper we revisit the classic multiview representation of 3D meshes and study several techniques that make them effective for 3D semantic segmentation of meshes. Given a 3D mesh reconstructed from RGBD sensors, our method effectively chooses different virtual views of the 3D mesh and renders multiple 2D c… ▽ More Semantic segmentation of 3D meshes is an important problem for 3D scene understanding. In this paper we revisit the classic multiview representation of 3D meshes and study several techniques that make them effective for 3D semantic segmentation of meshes. Given a 3D mesh reconstructed from RGBD sensors, our method effectively chooses different virtual views of the 3D mesh and renders multiple 2D channels for training an effective 2D semantic segmentation model. Features from multiple per view predictions are finally fused on 3D mesh vertices to predict mesh semantic segmentation labels. Using the large scale indoor 3D semantic segmentation benchmark of ScanNet, we show that our virtual views enable more effective training of 2D semantic segmentation networks than previous multiview approaches. When the 2D per pixel predictions are aggregated on 3D surfaces, our virtual multiview fusion method is able to achieve significantly better 3D semantic segmentation results compared to all prior multiview approaches and competitive with recent 3D convolution approaches. △ Less

Submitted 26 July, 2020; originally announced July 2020.

Comments: To appear in ECCV 2020

An LSTM Approach to Temporal 3D Object Detection in LiDAR Point Clouds

Authors: Rui Huang, Wanyue Zhang, Abhijit Kundu, Caroline Pantofaru, David A Ross, Thomas Funkhouser, Alireza Fathi

Abstract: Detecting objects in 3D LiDAR data is a core technology for autonomous driving and other robotics applications. Although LiDAR data is acquired over time, most of the 3D object detection algorithms propose object bounding boxes independently for each frame and neglect the useful information available in the temporal domain. To address this problem, in this paper we propose a sparse LSTM-based mult… ▽ More Detecting objects in 3D LiDAR data is a core technology for autonomous driving and other robotics applications. Although LiDAR data is acquired over time, most of the 3D object detection algorithms propose object bounding boxes independently for each frame and neglect the useful information available in the temporal domain. To address this problem, in this paper we propose a sparse LSTM-based multi-frame 3d object detection algorithm. We use a U-Net style 3D sparse convolution network to extract features for each frame's LiDAR point-cloud. These features are fed to the LSTM module together with the hidden and memory features from last frame to predict the 3d objects in the current frame as well as hidden and memory features that are passed to the next frame. Experiments on the Waymo Open Dataset show that our algorithm outperforms the traditional frame by frame approach by 7.5% [email protected] and other multi-frame approaches by 1.2% while using less memory and computation per frame. To the best of our knowledge, this is the first work to use an LSTM for 3D object detection in sparse point clouds. △ Less

Submitted 24 July, 2020; originally announced July 2020.

Comments: To appear in ECCV 2020

On stabilizability of switched linear systems under restricted switching

Authors: Atreyee Kundu

Abstract: This paper deals with stability of discrete-time switched linear systems whose all subsystems are unstable and the set of admissible switching signals obeys pre-specified restrictions on switches between the subsystems and dwell times on the subsystems. We derive sufficient conditions on the subsystems matrices such that a switched system is globally exponentially stable under a set of purely time… ▽ More This paper deals with stability of discrete-time switched linear systems whose all subsystems are unstable and the set of admissible switching signals obeys pre-specified restrictions on switches between the subsystems and dwell times on the subsystems. We derive sufficient conditions on the subsystems matrices such that a switched system is globally exponentially stable under a set of purely time-dependent switching signals that obeys the given restrictions. The main apparatuses for our analysis are (matrix) commutation relations between certain products of the subsystems matrices and graph-theoretic arguments. △ Less

Submitted 21 May, 2020; originally announced May 2020.

Comments: 20 pages, 2 figures

Journal ref: IEEE Transactions on Automatic Control, early access, 2021

Data-driven switching logic design for switched linear systems

Authors: Atreyee Kundu

Abstract: This paper deals with stabilization of discrete-time switched linear systems when explicit knowledge of the state-space models of their subsystems is not available. Given the set of admissible switches between the subsystems, the admissible dwell times on the subsystems and a set of finite traces of state trajectories of the subsystems that satisfies certain properties, we devise an algorithm that… ▽ More This paper deals with stabilization of discrete-time switched linear systems when explicit knowledge of the state-space models of their subsystems is not available. Given the set of admissible switches between the subsystems, the admissible dwell times on the subsystems and a set of finite traces of state trajectories of the subsystems that satisfies certain properties, we devise an algorithm that designs periodic switching logics which preserve stability of the resulting switched system. We combine two ingredients: (a) data-based stability analysis of discrete-time linear systems and (b) multiple Lyapunov-like functions and graph walks based design of stabilizing switching logics, for this purpose. A numerical example is presented to demonstrate the proposed algorithm. △ Less

Submitted 24 August, 2020; v1 submitted 11 March, 2020; originally announced March 2020.

Comments: 10 pages, 2 Figures, submitted for publication. arXiv admin note: text overlap with arXiv:2002.02087

Data-based computation of stabilizing minimum dwell times for discrete-time switched linear systems

Authors: Atreyee Kundu

Abstract: We present an algorithm to compute stabilizing minimum dwell times for discrete-time switched linear systems without the explicit knowledge of state-space models of their subsystems. Given a set of finite traces of state trajectories of the subsystems that satisfies certain properties, our algorithm involves the following tasks: first, multiple Lyapunov functions are designed from the given data;… ▽ More We present an algorithm to compute stabilizing minimum dwell times for discrete-time switched linear systems without the explicit knowledge of state-space models of their subsystems. Given a set of finite traces of state trajectories of the subsystems that satisfies certain properties, our algorithm involves the following tasks: first, multiple Lyapunov functions are designed from the given data; second, a set of relevant scalars is computed from these functions; and third, a stabilizing minimum dwell time is determined as a function of these scalars. A numerical example is presented to demonstrate the proposed algorithm. △ Less

Submitted 5 March, 2020; v1 submitted 5 February, 2020; originally announced February 2020.

Comments: 7 pages, 1 figure, submitted for conference publication

Scheduling networked control systems under jamming attacks

Authors: Atreyee Kundu

Abstract: This paper deals with the design of scheduling policies for networked control systems whose shared networks have limited communication capacity and the controller to plant channels are vulnerable to jamming attacks. We assume that among N plants, only M (< N) plants can communicate with their controllers at any time instant, and the attack sequences follow an (m,k)-firm model, i.e., in any k conse… ▽ More This paper deals with the design of scheduling policies for networked control systems whose shared networks have limited communication capacity and the controller to plant channels are vulnerable to jamming attacks. We assume that among N plants, only M (< N) plants can communicate with their controllers at any time instant, and the attack sequences follow an (m,k)-firm model, i.e., in any k consecutive time instants, the control inputs sent to some or all of the plants accessing the communication network, are deactivated at most at m (< k) time instants. We devise a new algorithm to allocate the network to the plants periodically such that stability of each plant is preserved under the admissible attack signals. The main apparatus for our analysis is a switched systems representation of the individual plants in an NCS. We rely on matrix commutators (Lie brackets) between the stable and unstable modes of operation of the plants to guarantee stability under our scheduling policies. △ Less

Submitted 5 February, 2020; originally announced February 2020.

Comments: 14 pages, 7 figures

Report number: A shorter version of this paper is accepted to IEEE Conference on Decision and Control 2020

Learning switched systems from simulation models

Authors: Atreyee Kundu

Abstract: The design of decision and control strategies for switched systems typically requires complete knowledge of (i) mathematical models of the subsystems and (ii) restrictions on admissible switches between the subsystems. We propose an active learning algorithm that infers (i) and (ii) for discrete-time switched systems whose subsystems dynamics are governed by sets of scalar polynomials and switchin… ▽ More The design of decision and control strategies for switched systems typically requires complete knowledge of (i) mathematical models of the subsystems and (ii) restrictions on admissible switches between the subsystems. We propose an active learning algorithm that infers (i) and (ii) for discrete-time switched systems whose subsystems dynamics are governed by sets of scalar polynomials and switching signals are constrained by automata. We collect data from gray-box simulation models of the switched systems for this purpose. Our technique for learning (i) involves linear algebraic tools, while for learning (ii) we employ a modified version of the well-known $L^*$-algorithm from machine learning literature. A numerical example is presented to demonstrate our learning algorithm. △ Less

Submitted 12 February, 2020; v1 submitted 28 November, 2019; originally announced November 2019.

Comments: 13 pages

Journal ref: A shorter version is at IEEE Conference on Decision and Control, 2021

On the design of stabilizing cycles for switched linear systems

Authors: Atreyee Kundu

Abstract: Given a family of systems, identifying stabilizing switching signals in terms of infinite walks constructed by concatenating cycles on the underlying directed graph of a switched system that satisfy certain conditions, is a well-known technique in the literature. This paper deals with a new {method to design} these cycles for stability of switched linear systems. We employ properties of the subsys… ▽ More Given a family of systems, identifying stabilizing switching signals in terms of infinite walks constructed by concatenating cycles on the underlying directed graph of a switched system that satisfy certain conditions, is a well-known technique in the literature. This paper deals with a new {method to design} these cycles for stability of switched linear systems. We employ properties of the subsystem matrices and mild assumption on the admissible switches between the subsystems {for this purpose}. In contrast to prior works, {our construction of} stabilizing cycles does not involve design of Lyapunov-like functions and storage of sets of scalars in memory prior to the application of a cycle detection algorithm. As a result, {the} techniques {proposed in this paper} offer improved numerical tractability. △ Less

Submitted 31 July, 2019; v1 submitted 1 June, 2019; originally announced June 2019.

Comments: 11 pages, 2 figures, To appear in IEEE Control Systems Letters

Journal ref: IEEE Control Systems Letters, vol. 4, no. 2, 2020, 385-390

Robust matrix commutator conditions for stability of switched linear systems under restricted switching

Authors: Atreyee Kundu, Debasish Chatterjee

Abstract: This article treats global uniform exponential stability (GUES) of discrete-time switched linear systems under restricted switching. Given admissible minimum and maximum dwell times, we provide sufficient conditions on the subsystems under which they admit a set of switching signals that obeys the given restrictions on dwell times and preserves stability of the resulting switched system. Our analy… ▽ More This article treats global uniform exponential stability (GUES) of discrete-time switched linear systems under restricted switching. Given admissible minimum and maximum dwell times, we provide sufficient conditions on the subsystems under which they admit a set of switching signals that obeys the given restrictions on dwell times and preserves stability of the resulting switched system. Our analysis relies on combinatorial arguments applied to matrix commutators and avoids the employment of Lyapunov-like functions. The proposed set of stabilizing switching signals is characterized in terms of duration of activation of Schur stable subsystems and non-consecutive activation of distinct unstable subsystems. △ Less

Submitted 15 July, 2019; v1 submitted 25 March, 2019; originally announced March 2019.

Journal ref: Automatica, vol. 115, 2020, article no.: 108904

A new condition for stability of switched linear systems under restricted minimum dwell time switching

Authors: Atreyee Kundu

Abstract: We propose matrix commutator based stability characterization for discrete-time switched linear systems under restricted switching. Given an admissible minimum dwell time, we identify sufficient conditions on subsystems such that a switched system is stable under all switching signals that obey the given restriction. The primary tool for our analysis is commutation relations between the subsystem… ▽ More We propose matrix commutator based stability characterization for discrete-time switched linear systems under restricted switching. Given an admissible minimum dwell time, we identify sufficient conditions on subsystems such that a switched system is stable under all switching signals that obey the given restriction. The primary tool for our analysis is commutation relations between the subsystem matrices. Our stability conditions are robust with respect to small perturbations in the elements of these matrices. In case of arbitrary switching (i.e., given minimum dwell time = 1), we recover the prior result [1,Proposition 1] as a special case of our result. △ Less

Submitted 29 November, 2019; v1 submitted 22 March, 2019; originally announced March 2019.

Comments: To appear in Systems and Control Letters

Journal ref: Systems and Control Letters, vol. 135, 2020, article no. 104597

Stabilizing Scheduling Policies for Networked Control Systems

Authors: Atreyee Kundu, Daniel E. Quevedo

Abstract: This paper deals with the problem of allocating communication resources for Networked Control Systems (NCSs). We consider an NCS consisting of a set of discrete-time LTI plants whose stabilizing feedback loops are closed through a shared communication channel. Due to a limited communication capacity of the channel, not all plants can exchange information with their controllers at any instant of ti… ▽ More This paper deals with the problem of allocating communication resources for Networked Control Systems (NCSs). We consider an NCS consisting of a set of discrete-time LTI plants whose stabilizing feedback loops are closed through a shared communication channel. Due to a limited communication capacity of the channel, not all plants can exchange information with their controllers at any instant of time. We propose a method to find periodic scheduling policies under which global asymptotic stability of each plant in the NCS is preserved. The individual plants are represented as switched systems, and the NCS is expressed as a weighted directed graph. We construct stabilizing scheduling policies by employing cycles on the underlying weighted directed graph of the NCS that satisfy appropriate contractivity conditions. We also discuss algorithmic design of these cycles. △ Less

Submitted 24 January, 2019; originally announced January 2019.

Journal ref: IEEE Transactions on Control of Network Systems, vol. 7, no. 1, 2020, pp. 163-175

On asymptotic characterization of destabilizing switching signals for switched linear systems

Authors: Atreyee Kundu

Abstract: This paper deals with classes of (de)stabilizing switching signals for switched systems. Most of the available conditions for stability of switched systems are sufficient in nature, and consequently, their violation does not conclude instability of a switched system. The study of instability is, however, important for obvious reasons. Our contributions are twofold: Firstly, we propose a class of s… ▽ More This paper deals with classes of (de)stabilizing switching signals for switched systems. Most of the available conditions for stability of switched systems are sufficient in nature, and consequently, their violation does not conclude instability of a switched system. The study of instability is, however, important for obvious reasons. Our contributions are twofold: Firstly, we propose a class of switching signals under which a continuous-time switched linear system is unstable. Our characterization of instability depends solely on the asymptotic behaviour of frequency of switching, frequency of transition between subsystems, and fraction of activation of subsystems. Secondly, we show that our class of destabilizing switching signals is a strict subset of the class of switching signals that does not satisfy asymptotic characterization of stability recently proposed in the literature. This observation identifies a gap between asymptotic characterizations of stabilizing and destabilizing switching signals for switched linear systems. The main apparatus for our analysis is multiple Lyapunov-like functions. △ Less

Submitted 22 December, 2018; originally announced December 2018.

Comments: In proceedings of the 6th Indian Control Conference, 2019

Stabilizing switching signals: a transition from point-wise to asymptotic conditions

Authors: Atreyee Kundu, Debasish Chatterjee

Abstract: Characterization of classes of switching signals that ensure stability of switched systems occupies a significant portion of the switched systems literature. This article collects a multitude of stabilizing switching signals under an umbrella framework. We achieve this in two steps: Firstly, given a family of systems, possibly containing unstable dynamics, we propose a new and general class of sta… ▽ More Characterization of classes of switching signals that ensure stability of switched systems occupies a significant portion of the switched systems literature. This article collects a multitude of stabilizing switching signals under an umbrella framework. We achieve this in two steps: Firstly, given a family of systems, possibly containing unstable dynamics, we propose a new and general class of stabilizing switching signals. Secondly, we demonstrate that prior results based on both point-wise and asymptotic characterizations follow our result. This is the first attempt in the switched systems literature where these switching signals are unified under one banner. △ Less

Submitted 22 May, 2017; v1 submitted 16 March, 2017; originally announced March 2017.

Comments: 7 pages

Journal ref: Systems & Control Letters, Vol. 106, pp. 16-23, 2017

A graph theoretic approach to input-to-state stability of switched systems

Authors: Atreyee Kundu, Debasish Chatterjee

Abstract: This article deals with input-to-state stability (ISS) of discrete-time switched systems. Given a family of nonlinear systems with exogenous inputs, we present a class of switching signals under which the resulting switched system is ISS. We allow non-ISS systems in the family and our analysis involves graph-theoretic arguments. A weighted digraph is associated to the switched system, and a switch… ▽ More This article deals with input-to-state stability (ISS) of discrete-time switched systems. Given a family of nonlinear systems with exogenous inputs, we present a class of switching signals under which the resulting switched system is ISS. We allow non-ISS systems in the family and our analysis involves graph-theoretic arguments. A weighted digraph is associated to the switched system, and a switching signal is expressed as an infinite walk on this digraph, both in a natural way. Our class of stabilizing switching signals (infinite walks) is periodic in nature and affords simple algorithmic construction. △ Less

Submitted 9 September, 2015; originally announced September 2015.

Comments: 14 pages, 1 figure

Journal ref: European Journal of Control, Vol. 29, No. 5, pp. 44-50, 2016

Generalized switching signals for input-to-state stability of switched systems

Authors: Atreyee Kundu, Debasish Chatterjee, Daniel Liberzon

Abstract: This article deals with input-to-state stability (ISS) of continuous-time switched nonlinear systems. Given a family of systems with exogenous inputs such that not all systems in the family are ISS, we characterize a new and general class of switching signals under which the resulting switched system is ISS. Our stabilizing switching signals allow the number of switches to grow faster than an affi… ▽ More This article deals with input-to-state stability (ISS) of continuous-time switched nonlinear systems. Given a family of systems with exogenous inputs such that not all systems in the family are ISS, we characterize a new and general class of switching signals under which the resulting switched system is ISS. Our stabilizing switching signals allow the number of switches to grow faster than an affine function of the length of a time interval, unlike in the case of average dwell time switching. We also recast a subclass of average dwell time switching signals in our setting and establish analogs of two representative prior results. △ Less

Submitted 9 February, 2015; originally announced February 2015.

Comments: 16 pages, 2 figures

Journal ref: Automatica, Vol. 64, No. 2, pp. 270-277, 2016

Deterministic and probabilistic algorithms for stabilizing discrete-time switched linear systems

Authors: Atreyee Kundu, Niranjan Balachandran, Debasish Chatterjee

Abstract: In this article we study algorithmic synthesis of the class of stabilizing switching signals for discrete-time switched linear systems proposed in [12]. A weighted digraph is associated in a natural way to a switched system, and the switching signal is expressed as an infinite walk on this weighted digraph. We employ graph-theoretic tools and discuss different algorithms for designing walks whose… ▽ More In this article we study algorithmic synthesis of the class of stabilizing switching signals for discrete-time switched linear systems proposed in [12]. A weighted digraph is associated in a natural way to a switched system, and the switching signal is expressed as an infinite walk on this weighted digraph. We employ graph-theoretic tools and discuss different algorithms for designing walks whose corresponding switching signals satisfy the stabilizing switching conditions proposed in [12]. We also address the issue of how likely/generic it is for a family of systems to admit stabilizing switching signals, and under mild assumptions give sufficient conditions for the same. Our solutions have both deterministic and probabilistic flavours. △ Less

Submitted 17 September, 2014; v1 submitted 8 May, 2014; originally announced May 2014.

Comments: 11 pages, 2 figures

MSC Class: 93C30; 93D20

Stabilizing discrete-time switched linear systems

Authors: Atreyee Kundu, Debasish Chatterjee

Abstract: This article deals with stabilizing discrete-time switched linear systems. Our contributions are threefold: Firstly, given a family of linear systems possibly containing unstable dynamics, we propose a large class of switching signals that stabilize a switched system generated by the switching signal and the given family of systems. Secondly, given a switched system, a sufficient condition for the… ▽ More This article deals with stabilizing discrete-time switched linear systems. Our contributions are threefold: Firstly, given a family of linear systems possibly containing unstable dynamics, we propose a large class of switching signals that stabilize a switched system generated by the switching signal and the given family of systems. Secondly, given a switched system, a sufficient condition for the existence of the proposed switching signal is derived by expressing the switching signal as an infinite walk on a directed graph representing the switched system. Thirdly, given a family of linear systems, we propose an algorithmic technique to design a switching signal for stabilizing the corresponding switched system. △ Less

Submitted 14 October, 2013; originally announced October 2013.

Comments: 10 pages

MSC Class: 93D20

Stabilizing switching signals for switched linear systems

Authors: Atreyee Kundu, Debasish Chatterjee

Abstract: This article deals with stability of continuous-time switched linear systems under constrained switching. Given a family of linear systems, possibly containing unstable dynamics, we characterize a new class of switching signals under which the switched linear system generated by it and the family of systems is globally asymptotically stable. Our characterization of such stabilizing switching signa… ▽ More This article deals with stability of continuous-time switched linear systems under constrained switching. Given a family of linear systems, possibly containing unstable dynamics, we characterize a new class of switching signals under which the switched linear system generated by it and the family of systems is globally asymptotically stable. Our characterization of such stabilizing switching signals involves the asymptotic frequency of switching, the asymptotic fraction of activation of the constituent systems, and the asymptotic densities of admissible transitions among them. Our techniques employ multiple Lyapunov-like functions, and extend preceding results both in scope and applicability. △ Less

Submitted 22 March, 2013; v1 submitted 6 March, 2013; originally announced March 2013.

Comments: 15 pages; corrected version with new references

MSC Class: 93D20