arXiv subjects

Vidyasagar Sadhu

Publications and source records attributed to Vidyasagar Sadhu.

At least 19 recordsLinked to original sources

arXiv2025

ToMCAT: Theory-of-Mind for Cooperative Agents in Teams via Multiagent Diffusion Policies

In this paper we present ToMCAT (Theory-of-Mind for Cooperative Agents in Teams), a new framework for generating ToM-conditioned trajectories. It combines a meta-learning mechanism, that performs ToM reasoning over teammates' underlying goals and future behavior, with a multiagent denoising-diffusion model, that generates plans for an agent and its teammates conditioned on both the agent's goals and its teammates' characteristics, as computed via ToM. We implemented an online planning system that dynamically samples new trajectories (replans) from the diffusion model whenever it detects a divergence between a previously generated plan and the current state of the world. We conducted several experiments using ToMCAT in a simulated cooking domain. Our results highlight the importance of the dynamic replanning mechanism in reducing the usage of resources without sacrificing team performance. We also show that recent observations about the world and teammates' behavior collected by an agent over the course of an episode combined with ToM inferences are crucial to generate team-aware plans for dynamic adaptation to teammates, especially when no prior information is provided about them.

cs.MA↗

arXiv2025

Real-Time Navigation for Autonomous Aerial Vehicles Using Video

Most applications in autonomous navigation using mounted cameras rely on the construction and processing of geometric 3D point clouds, which is an expensive process. However, there is another simpler way to make a space navigable quickly: to use semantic information (e.g., traffic signs) to guide the agent. However, detecting and acting on semantic information involves Computer Vision~(CV) algorithms such as object detection, which themselves are demanding for agents such as aerial drones with limited onboard resources. To solve this problem, we introduce a novel Markov Decision Process~(MDP) framework to reduce the workload of these CV approaches. We apply our proposed framework to both feature-based and neural-network-based object-detection tasks, using open-loop and closed-loop simulations as well as hardware-in-the-loop emulations. These holistic tests show significant benefits in energy consumption and speed with only a limited loss in accuracy compared to models based on static features and neural networks.

cs.RO↗

arXiv2023

Sensor Control for Information Gain in Dynamic, Sparse and Partially Observed Environments

We present an approach for autonomous sensor control for information gathering under partially observable, dynamic and sparsely sampled environments that maximizes information about entities present in that space. We describe our approach for the task of Radio-Frequency (RF) spectrum monitoring, where the goal is to search for and track unknown, dynamic signals in the environment. To this end, we extend the Deep Anticipatory Network (DAN) Reinforcement Learning (RL) framework by (1) improving exploration in sparse, non-stationary environments using a novel information gain reward, and (2) scaling up the control space and enabling the monitoring of complex, dynamic activity patterns using hybrid convolutional-recurrent neural layers. We also extend this problem to situations in which sampling from the intended RF spectrum/field is limited and propose a model-based version of the original RL algorithm that fine-tunes the controller via a model that is iteratively improved from the limited field sampling. Results in simulated RF environments of differing complexity show that our system outperforms the standard DAN architecture and is more flexible and robust than baseline expert-designed agents. We also show that it is adaptable to non-stationary emission environments.

cs.AI↗

arXiv2023

Automatic Measures for Evaluating Generative Design Methods for Architects

The recent explosion of high-quality image-to-image methods has prompted interest in applying image-to-image methods towards artistic and design tasks. Of interest for architects is to use these methods to generate design proposals from conceptual sketches, usually hand-drawn sketches that are quickly developed and can embody a design intent. More specifically, instantiating a sketch into a visual that can be used to elicit client feedback is typically a time consuming task, and being able to speed up this iteration time is important. While the body of work in generative methods has been impressive, there has been a mismatch between the quality measures used to evaluate the outputs of these systems and the actual expectations of architects. In particular, most recent image-based works place an emphasis on realism of generated images. While important, this is one of several criteria architects look for. In this work, we describe the expectations architects have for design proposals from conceptual sketches, and identify corresponding automated metrics from the literature. We then evaluate several image-to-image generative methods that may address these criteria and examine their performance across these metrics. From these results, we identify certain challenges with hand-drawn conceptual sketches and describe possible future avenues of investigation to address them.

cs.CV↗

arXiv2020

On-board Deep-learning-based Unmanned Aerial Vehicle Fault Cause Detection and Identification

With the increase in use of Unmanned Aerial Vehicles (UAVs)/drones, it is important to detect and identify causes of failure in real time for proper recovery from a potential crash-like scenario or post incident forensics analysis. The cause of crash could be either a fault in the sensor/actuator system, a physical damage/attack, or a cyber attack on the drone's software. In this paper, we propose novel architectures based on deep Convolutional and Long Short-Term Memory Neural Networks (CNNs and LSTMs) to detect (via Autoencoder) and classify drone mis-operations based on sensor data. The proposed architectures are able to learn high-level features automatically from the raw sensor data and learn the spatial and temporal dynamics in the sensor data. We validate the proposed deep-learning architectures via simulations and experiments on a real drone. Empirical results show that our solution is able to detect with over 90% accuracy and classify various types of drone mis-operations (with about 99% accuracy (simulation data) and upto 88% accuracy (experimental data)).

eess.SP↗

arXiv2020

Energy-efficient Analog Sensing for Large-scale and High-density Persistent Wireless Monitoring

The research challenge of current Wireless Sensor Networks (WSNs) is to design energy-efficient, low-cost, high-accuracy, self-healing, and scalable systems for applications such as environmental monitoring. Traditional WSNs consist of low density, power-hungry digital motes that are expensive and cannot remain functional for long periods on a single power charge. In order to address these challenges, a dumb-sensing and smart-processing architecture that splits sensing and computation capabilities is proposed. Sensing is exclusively the responsibility of analog substrate---consisting of low-power, low-cost all-analog sensors---that sits beneath the traditional WSN comprising of digital nodes, which does all the processing of the sensor data received from analog sensors. A low-power and low-cost solution for substrate sensors has been proposed using Analog Joint Source Channel Coding (AJSCC) realized via the characteristics of Metal Oxide Semiconductor Field Effect Transistor (MOSFET). Digital nodes (receiver) also estimate the source distribution at the analog sensors (transmitter) using machine learning techniques so as to find the optimal parameters of AJSCC that are communicated back to the analog sensors to adapt their sensing resolution as per the application needs. The proposed techniques have been validated via simulations from MATLAB and LTSpice to show promising performance and indeed prove that our framework can support large scale high density and persistent WSN deployment.

eess.SP↗

arXiv2020

Signal Recovery Performance Analysis in Wireless Sensing with Rectangular-Type Analog Joint Source-Channel Coding

The signal recovery performance of the rectangular-type Analog Joint Source-Channel Coding (AJSCC) is analyzed in this work for high and medium/low Signal-to-Noise Ratio (SNR) scenarios in the wireless sensing systems. The analysis and derivations of the medium/low SNR scenario are based on the comprehensive listing of all the signal variation cases in the three-dimensional signal mapping curve of the rectangular-type AJSCC. Theoretical formulations of Mean Square Error (MSE) performance are derived for both analog sensing and digital sensing systems with rectangular-type AJSCC. Evaluation results indicate that, there are optimal parameters in the rectangular-type AJSCC to minimize the signal recovery MSE performance at high and medium/low SNR scenarios. In addition, the performance of digital sensing with low-resolution Analog-to-Digital Conversion (ADC) is compared with analog sensing for both high and medium/low SNR scenarios in this work. The theoretical and evaluation results have practical value to the wireless sensing system designs based on the rectangular-type AJSCC.

eess.SP↗

arXiv2019

Real-time Image Enhancement for Vision-based Autonomous Underwater Vehicle Navigation in Murky Waters

Classic vision-based navigation solutions, which are utilized in algorithms such as Simultaneous Localization and Mapping (SLAM), usually fail to work underwater when the water is murky and the quality of the recorded images is low. That is because most SLAM algorithms are feature-based techniques and often it is impossible to extract the matched features from blurry underwater images. To get more useful features, image processing techniques can be used to dehaze the images before they are used in a navigation/localization algorithm. There are many well-developed methods for image restoration, but the degree of enhancement and the resource cost of the methods are different. In this paper, we propose a new visual SLAM, specifically-designed for the underwater environment, using Generative Adversarial Networks (GANs) to enhance the quality of underwater images with underwater image quality evaluation metrics. This procedure increases the efficiency of SLAM and gets a better navigation and localization accuracy. We evaluate the proposed GANs-SLAM combination by using different images with various levels of turbidity in the water. Experiments were conducted and the data was extracted from the Carnegie Lake in Princeton, and the Raritan river both in New Jersey, USA.

eess.IV↗

arXiv2019

UW-MARL: Multi-Agent Reinforcement Learning for Underwater Adaptive Sampling using Autonomous Vehicles

Near-real-time water-quality monitoring in uncertain environments such as rivers, lakes, and water reservoirs of different variables is critical to protect the aquatic life and to prevent further propagation of the potential pollution in the water. In order to measure the physical values in a region of interest, adaptive sampling is helpful as an energy- and time-efficient technique since an exhaustive search of an area is not feasible with a single vehicle. We propose an adaptive sampling algorithm using multiple autonomous vehicles, which are well-trained, as agents, in a Multi-Agent Reinforcement Learning (MARL) framework to make efficient sequence of decisions on the adaptive sampling procedure. The proposed solution is evaluated using experimental data, which is fed into a simulation framework. Experiments were conducted in the Raritan River, Somerset and in Carnegie Lake, Princeton, NJ during July 2019.

eess.SY↗

arXiv2019

Low-power All-analog Circuit for Rectangular-type Analog Joint Source Channel Coding

A low-complexity all-analog circuit is proposed to perform efficiently Analog Joint Source Channel Coding (AJSCC), which can compress two or more sensor signals into one with controlled distortion while also being robust against wireless channel impairments. The idea is to realize the rectangular-type AJSCC using Voltage Controlled Voltage Sources (VCVS). The proposal is verified by Spice simulations as well as breadboard and Printed Circuit Board (PCB) implementations. Results indicate that the design is feasible for low-complexity systems like persistent wireless sensor networks requiring low circuit power.

eess.SP↗

arXiv2019

Towards Low-power Wearable Wireless Sensors for Molecular Biomarker and Physiological Signal Monitoring

A low-power wearable wireless sensor measuring both molecular biomarkers and physiological signals is proposed, where the former are measured by a microfluidic biosensing system while the latter are measured electrically. The low-power consumption of the sensor is achieved by an all-analog circuit implementing Analog Joint Source-Channel Coding (AJSCC) compression. The sensor is applicable to a wide range of biomedical applications that require real-time concurrent molecular biomarker and physiological signal monitoring.

eess.SP↗

arXiv2019

Towards Ultra-low-power Realization of Analog Joint Source-Channel Coding using MOSFETs

Certain sensing applications such as Internet of Things (IoTs), where the sensing phenomenon may change rapidly in both time and space, requires sensors that consume ultra-low power (so that they do not need to be put to sleep leading to loss of temporal and spatial resolution) and have low costs (for high density deployment). A novel encoding based on Metal Oxide Semiconductor Field Effect Transistors (MOSFETs) is proposed to realize Analog Joint Source Channel Coding (AJSCC), a low-complexity technique to compress two (or more) signals into one with controlled distortion. In AJSCC, the y-axis is quantized while the x-axis is continuously captured. A power-efficient design to support multiple quantization levels is presented so that the digital receiver can decide the optimum quantization and the analog transmitter circuit is able to realize that. The approach is verified via Spice and MATLAB simulations.

cs.ET↗

arXiv2019

Improved Circuit Design of Analog Joint Source Channel Coding for Low-power and Low-complexity Wireless Sensors

To enable low-power and low-complexity wireless monitoring, an improved circuit design of Analog Joint Source Channel Coding (AJSCC) is proposed for wireless sensor nodes. This innovative design is based on Analog Divider Blocks (ADB) with tunable spacing between AJSCC levels. The ADB controls the switching between two types of Voltage Controlled Voltage Sources (VCVS). LTSpice simulations were performed to evaluate the performance of the circuit, and the power consumption and circuit complexity of this new ADB-based design were compared with our previous parallel-VCVS design. It is found that this improved circuit design based on ADB outperforms the design based on parallel VCVS for a large number of AJSCC levels (>= 16), both in terms of power consumption as well as circuit complexity, thus enabling persistent and higher temporal/spatial resolution environmental sensing.

eess.SP↗

arXiv2019

Analog Signal Compression and Multiplexing Techniques for Healthcare Internet of Things

Scalability is a major issue for Internet of Things (IoT) as the total amount of traffic data collected and/or the number of sensors deployed grow. In some IoT applications such as healthcare, power consumption is also a key design factor for the IoT devices. In this paper, a multi-signal compression and encoding method based on Analog Joint Source Channel Coding (AJSCC) is proposed that works fully in the analog domain without the need for power-hungry Analog-to-Digital Converters (ADCs). Compression is achieved by quantizing all the input signals but one. While saving power, this method can also reduce the number of devices by combining one or more sensing functionalities into a single device (called 'AJSCC device'). Apart from analog encoding, AJSCC devices communicate to an aggregator node (FPMM receiver) using a novel Frequency Position Modulation and Multiplexing (FPMM) technique. Such joint modulation and multiplexing technique presents three mayor advantages---it is robust to interference at particular frequency bands, it protects against eavesdropping, and it consumes low power due to a very low Signal-to-Noise Ratio (SNR) operating region at the receiver. Performance of the proposed multi-signal compression method and FPMM technique is evaluated via simulations in terms of Mean Square Error (MSE) and Miss Detection Rate (MDR), respectively.

eess.SP↗

arXiv2019

Towards Wireless Health Monitoring via Analog Signal Compression based Biosensing Platform

Wireless all-analog biosensor design for concurrent microfluidic and physiological signal monitoring is presented in this work. The key component is an all-analog circuit capable of compressing two analog sources into one analog signal by Analog Joint Source-Channel Coding (AJSCC). Two circuit designs are discussed, including the stacked-Voltage Controlled Voltage Source (VCVS) design with the fixed number of levels, and an improved design, which supports a flexible number of AJSCC levels. Experimental results are presented on the wireless biosensor prototype, composed of Printed Circuit Board (PCB) realizations of the stacked-VCVS design. Furthermore, circuit simulation and wireless link simulation results are presented on the improved design. Results indicate that the proposed wireless biosensor is well suited for sensing two biological signals simultaneously with high accuracy, and can be applied to a wide variety of low-power and low-cost wireless continuous health monitoring applications.

eess.SP↗

arXiv2019

Secure Mobile Technologies for Proactive Critical Infrastructure Situational Awareness

Trustworthy operation of our national critical infrastructures, such as the electricity grid, against adversarial parties and accidental failures requires constant and secure monitoring capabilities. In this paper, Eyephone is presented to leverage secure smartphone sensing and data acquisition capabilities and enable pervasive sensing of the national critical infrastructures. The reported information by the smartphone users will notify the control center operators about particular accidental or malicious remote critical infrastructure incidents. The reporting will be proactive regarding potentially upcoming failures given the system's current risky situation, e.g., a tree close to fall on a power grid transmission line. The information will include various modalities such as images, video, audio, time and location. Eyephone will use system-wide information flow analysis and policy enforcement to prevent user privacy violations during the incident reportings. A working proof-of-concept prototype of Eyephone is implemented. Our results show that Eyephone allows secure and effective use of smartphones for real-time situational awareness of our national critical infrastructures.

cs.CR↗

arXiv2019

Energy-efficient Wireless Analog Sensing for Persistent Underwater Environmental Monitoring

The design of sensors or "things" as part of the new Internet of Underwater Things (IoUTs) paradigm comes with multiple challenges including limited battery capacity, not polluting the water body, and the ability to track continuously phenomena with high temporal/spatial variability. We claim that traditional digital sensors are incapable to meet these demands because of their high power consumption, high complexity (cost), and the use of non-biodegradable materials. To address the above challenges, we propose a novel architecture consisting of a sensing substrate of dense analog biodegradable sensors over which lies the traditional Wireless Sensor Network (WSN). The substrate analog biodegradable sensors perform Shannon mapping (a data-compression technique) using just a single Field Effect Transistor (FET) without the need for power-hungry Analog-to-Digital Converters (ADCs) resulting in much lower power consumption, complexity, and the ability to be powered using only sustainable energy-harvesting techniques. A novel and efficient decoding technique is also presented. Both encoding/decoding techniques have been verified via Spice and MATLAB simulations accounting for underwater acoustic channel variations.

eess.SP↗

arXiv2019

Deep Multi-Task Learning for Anomalous Driving Detection Using CAN Bus Scalar Sensor Data

Corner cases are the main bottlenecks when applying Artificial Intelligence (AI) systems to safety-critical applications. An AI system should be intelligent enough to detect such situations so that system developers can prepare for subsequent planning. In this paper, we propose semi-supervised anomaly detection considering the imbalance of normal situations. In particular, driving data consists of multiple positive/normal situations (e.g., right turn, going straight), some of which (e.g., U-turn) could be as rare as anomalous situations. Existing machine learning based anomaly detection approaches do not fare sufficiently well when applied to such imbalanced data. In this paper, we present a novel multi-task learning based approach that leverages domain-knowledge (maneuver labels) for anomaly detection in driving data. We evaluate the proposed approach both quantitatively and qualitatively on 150 hours of real-world driving data and show improved performance over baseline approaches.

cs.LG↗