Abstract

The objective of this study is to examine the interdependence between scene representation and robust autonomy of mobile robots such as quadrotors and cars. It addresses crucial issues with the motion planning and SLAM components in the navigation system. Voxel Grids built from range sensors have been a popular choice of world representation for motion planning. The planners query the map for distance to collision and gradients, the planner uses these measurements to generate a smooth and safe trajectory through trajectory optimization. However, these methods assume the presence of zero sensor noise, such an assumption is severely violated when the maps are built through RGBD cameras, the noise in the point clouds are transferred to the occupancy maps through a non-linear transform and is further propagated to the distance and gradient field which results in a non-parametric noise. The performance of the deterministic planners that rely on such inaccurate gradients was found to decrease. As an antidote we propose CCO-VOXEL the very first chance-constrained optimization (CCO) algorithm that can compute trajectory plans with probabilistic safety guarantees in real-time directly on the voxel-grid representation of the world. We leverage the notion of Hilbert Space embedding of distributions and Maximum Mean Discrepancy (MMD) and gradient free Cross-Entropy Method for obtaining its minimum. We also show how a combination of low-dimensional feature embedding and pre-caching of Kernel Matrices of MMD allows us to achieve real-time performance in simulations as well as in implementations on on-board commodity hardware that controls the quadrotor flight. Voxel maps are not robust to outliers and fail to capture the world geometry when built from a highly noisy and sparse point cloud, such as those generated from triangulation through monocular visualinertial SLAM (VI-SLAM). Inaccurate maps like these can significantly decrease the performance of the motion planner. We propose UrbanFly , a quadrotor planning framework for navigation in a planar urban high-rise environment. The core importance of UrbanFly is its ability to handle sparse and noisy point clouds built from VI-SLAM. Based on the insight that the sky-scrappers are planar, we derived an Uncertainty-Integrated Cuboidal (UIC) representation of the world and developed an uncertaintyaware planner that can interpret and leverage the benefits of UIC to generate safe trajectories in real time. UrbanFly demonstrates a 3.5 times improvement in safety metrics compared to state-of- the-art algorithms. Points clouds generated from LiDAR need a lot of memory to store and transfer and lack structure, making it hard to extract spatial features. This makes it challenging to deploy lidar based Loop Detection and Closure (LDC) for distributed collaborative SLAM. To address this, we present FinderNet, a 6-DOF vi vii distributed LDC system. We develop a novel DEM representation to reveal the underlying geometric structure of the point clouds, and an auto-encoder decoder structure to compress the point clouds. LDC is performed in the compressed space, allowing deployment in a multi-agent setting. Our approach provides view-invariance through a process of canonicalization and differentiable alignment, which is different from existing methods that rely on feature aggregation or overlap estimation. FinderNet has been evaluated on real-world datasets such as Kitti, GPR, and Oxford Robot Car.

Abstract

In this work, we propose IndoLayout, a novel real- time approach for generating high-quality occupancy maps from an RGB image for indoor scenes. Such occupancy maps are often crucial for pathplanning and mapping in indoor environments but are often built using only information contained in the ego view. In contrast, our approach also predicts occupancy values beyond immediately visible regions from just a monocular image, leveraging learnt priors from indoor scenes. Hence, our proposed network can produce a hallucinated, amodal scene layout that includes areas occluded in the RGB image, such as a navigable floor behind a desk. Specifically, we propose a novel architecture that uses self-attention and adversarial learning to vastly improve the quality of the predicted layout. We evaluate our model on several photorealistic indoor datasets and outperform previous relevant work on all metrics that measure layout quality, including newly adopted ones. Finally, we demonstrate the effectiveness of our method by showing significant improvements on the Point Goal navigation task over similar approaches using IndoLayout.

Abstract

Social media has experienced significant growth in the past decade and has enabled people to connect with people all over the world, have increased access to information and have an opportunity to express themselves and join like-minded communities. However, hate speech and online harassment are significant problems, with around two-thirds of adults under 30 having experienced some form of online harassment. Therefore, it becomes essential to regulate content on social media and it needs to be done automatically due to the large volume of daily content. In this thesis, we attempt to solve the above-mentioned problems by building best-in-class classifiers for novel datasets. One way to tackle harmful content online is to have a positive reinforcement approach and encourage positive and supportive messages. We propose a Hope Speech Detection model trained on a first-of-a-kind hope speech dataset. In the first approach, we used contextual embeddings to train classifiers using logistic regression, random forest, SVM, and LSTM based models. The second approach used a majority voting ensemble of 11 models obtained by fine-tuning pre-trained transformer models. Our model ranks first in terms of F1 score in the English language. While supporting and boosting positive content online is helpful, there should also be a distinction made between content that is positive and content that seems positive but encourages emotion suppression. Over the past few years, there has been a growing concern around toxic positivity on social media, a phenomenon where positivity is used to minimize one’s emotional experience. In this thesis, we create a dataset for toxic positivity classification from Twitter and an inspirational quote website. We then perform benchmarking experiments using various text classification models and show the suitability of these models for the task. While there are many hate speech classifiers trained on a generic hate speech definition, there is a lack of datasets that focus on homophobia and transphobia. In this thesis, we describe our approach to classify homophobia and transphobia in social media comments. We used an ensemble of transformer based models to build our classifier. Our classifier ranks 1st in terms of F1 score

Abstract

The discovery of new molecules and materials helps expand the horizons of novel and innovative real-life applications. In the pursuit of finding molecules with desired properties, chemists have traditionally relied on experimentation and recently on combinatorial methods to generate new substances often complimented by computational methods. The sheer size of the chemical space makes it infeasible to search through all possible molecules exhaustively. This calls for fast and efficient methods to navigate the chemical space to find substances with desired properties. This class of problems is referred to as inverse design problems. There is a variety of inverse problems in chemistry encompassing various subfields like drug discovery, retrosynthesis, structure identification etc. Recent developments in modern machine learning (ML) methods have shown great promise in being able to tackle problems of this kind. This has helped in making major strides in all key phases of molecule discovery ranging from in silico candidate generation to their synthesis with focus on small organic molecules. Optimization techniques like Bayesian optimization, reinforcement learning, attention-based transformers, deep generative models like variational autoencoders and generative adversarial networks form a robust arsenal of methods. The first chapter of this thesis summarizes the development of deep learning to tackle a wide variety of inverse design problems in chemistry towards the quest for synthesizing small organic compounds with purpose. Spectroscopy is the study of how matter interacts with electromagnetic radiations of specific frequencies that has led to several monumental discoveries in science. The spectra of any particular molecule is highly information-rich; while structure to spectra is straightforward using computational methods, the inverse relation of spectra to the corresponding molecular structure is still an unsolved problem. Nuclear Magnetic Resonance (NMR) spectroscopy is one such critical technique in the scientists’ toolkit to characterise small organic molecules to biomolecular structures like proteins and nucleic acids. In the second half of the thesis, a novel machine learning framework is proposed that attempts to solve this inverse problem by navigating the chemical space to find the correct structure given an NMR spectra. The proposed framework uses a combination of online Monte-Carlo-Tree-Search (MCTS) and a set of offline trained Graph Convolution Networks to build a molecule iteratively from scratch. Our method is able to predict the correct structure of the molecule ∼ 80% of the time in its top 3 guesses. We believe that the proposed framework is a significant step in solving the inverse design problem of NMR spectra to molecule that would be a significant step forward in high-throughput molecular synthesis

Abstract

In the recent years we have seen an enormous increase in the utilisation of Artificial Intelligence (AI), and the healthcare sector has not been exempt from this transformation. Machine learning systems have demonstrated unmatched success in the healthcare sector thanks to recent developments in digitalization and the massive influx of biomedical data. This could be a game changer for the sector. The rise of cardiovascular diseases across the globe has made electrocardiograms (ECGs) a crucial modality for diagnosis, owing to their non-invasive nature and simplicity. However, gathering 12-lead ECG data is an arduous task outside clinical settings. Wearables can collect an ECG with fewer leads than the standard 12 or the Photoplethysmogram (PPG) data, but medical professionals and conventional ECG analysis software find this data challenging to interpret. To address this issue, ECG reconstruction has been proposed. The second chapter provides a summary of the current applications of AI in diagnosis, prognosis, and therapy, along with its implications for combating the COVID-19 pandemic. The chapter also identifies obstacles to AI’s widespread adoption in the healthcare sector and suggests remedies to help usher in a more intelligent medical future. A unique single-lead to multi-lead ECG reconstruction method is suggested in the third chapter of this thesis employing a modified Attention U-Net framework. Our model, which was solely trained on lead II of ECG, is capable of replicating the remaining 11 leads of the conventional 12-lead ECG with a Pearson correlation of 0.805, a mean square error of 0.0122, and an R-squared value of 0.639. Moreover, a single combined model is used to reconstruct all 11 leads simultaneously, improving performance and simultaneously reducing the computational resources needed for training compared to current literature in the field. The model’s ability for real-life use was also demonstrated by training a deep learning model for multi-disease classification using actual 12-lead ECG data and testing it on both original and reconstructed 12-lead ECG signals. Comparable classification accuracies for both original and reconstructed signals suggest that the proposed model can preserve diagnostically relevant artefacts. The fourth chapter proposes a new approach using a Wasserstein generative adversarial network to convert PPG signals into single lead ECG signals. Unlike previous studies, we utilize longer 3-second PPG segments to generate a complete 3-second ECG signal in a single step. We also address a significant issue in earlier studies where same patient’s data was used for both model training and testing. To address this limitation, patient-specific data segmentation for the train and test set was employed to obtain more reliable results. In summary, this thesis proposes solutions for multi-lead ECG reconstruction and PPG to ECG conversion, aiming to bridge the divide between easily collected heart monitoring data and easily interpretable heart monitoring data.

Abstract

When the Indian government declared the first lockdown on 25 March 2020 to control the increasing number of COVID-19 cases, people were forced to stay and work from home. The aim of this study is to quantify the impact of stay-at-home orders on residential Air Conditioning (AC) energy and household electricity consumption (excluding AC energy). This was done using monitored data from 380 homes in a group of five buildings in Hyderabad, India. We gathered AC energy and household electricity consumption data at a 30-minute interval for each home individually in April 2019 and April 2020. Descriptive and inferential statistical analysis was done on this data. To compensate the difference in temperatures for the month of April in 2019 and 2020, only those weekdays were selected where the average temperature in 2019 was same as the average temperature in 2020. The study establishes that the average number of hours the AC was used per day in each home increased in the range 4.90 – 7.45% depending on the temperature for the year 2020. This can be expected as occupants stayed at home and they have used AC more frequently during lockdown period. Correspondingly, the overall AC consumption increased in the range 3.60 – 4.5%, however the daytime (8:00 AM to 8:00 PM) AC energy consumption increased in the range 22 – 26% and nighttime (8:00 PM to 8:00 AM) AC energy consumption decreased by 5-7% in the year 2020. As the AC was used throughout the day during lockdown period, due to pre-cooling effect the AC consumption during the night time got reduced. The study showed a rise in household daily electricity consumption of about 15% for the year 2020. The household electricity consumption increased during daytime by 22- 27.50% and 1.90- 6.6% during the nighttime. It can be explained as occupants worked from home, and used monitors, laptops, T.V, internet, lighting which in turn increased the household electricity consumption. It was observed that the morning household electricity peak demand shifted from 7:00 AM in 2019 to 9:00 AM in 2020. Conversely, the evening peak demand shifted from 9:00 PM in 2019 to 7:00 PM in 2020. An additional peak was observed during afternoon hours in the lockdown. This indicates the occupants started waking up late, as they didn’t have to commute to their workplaces. Also, there is an earlier evening peak for 2020 than 2019, as occupants in 2019 used to come home after 6:00 PM from their workplaces. Further, study was conducted for two sets of temperatures (Set_lo and Set_hi) to study the impact of temperature within the years 2019 and 2020. The study establishes that the AC was used more (4.80- 5.60%) for the higher temperature days when compared within the year. As expected the AC consumption is more for the higher temperature range. Furthermore, the consumption pattern throughout the day is similar for both the sets each year. Similarly, the average number of hours the AC was used per day in each home increased by 3.40 – 6.00% for the higher temperature range within the year. The study showed that household electricity consumption was more by 3.00% for the lower temperature range when compared higher temperature range. However, the household electricity consumption pattern is similar for both sets of temperatures

Abstract

Path planning for autonomous systems is a fundamental problem in robotics, especially when the task assigned to a robot is heavily dependent on the motion of one or more of its constituent parts. Efficient path planning requires fast adaptation to changes in the environment and generalizability to a variety of tasks while being intricately linked with a motion planner that handles the kinematic and dynamic constraints of the robot and its workspace. Present-day planning algorithms for robots fail to achieve real-time performance for robot systems with complex kinematics and heavy interactions such as multi-robot systems and robot manipulators. In fact, the motion planning objectives for these complex systems often present mathematical infeasibilities when posed as an optimization problem or become computationally cumbersome to compute by pure sampling. Further, these algorithms do not generalize to the task of collision avoidance against different types of obstacles for multi-robot systems or different types of end-effectors in robots with high-dimensional articulation. The primary focus of this dissertation is to present computationally-tractable solutions to the trajectory optimization problem for both multi-robot systems and manipulators. We explore gradient-based and stochastic optimization techniques and perform mathematical reformulations to adapt them to a wide variety of applications. First, we provide the requisite background in robot path planning and motion planning, including robot kinematics, trajectory representation methods, and collision-avoidance techniques. We also discuss state-of-the-art methods in gradient-based trajectory optimization and stochastic trajectory optimization. Next, we present a distributed GPU-based multi-agent trajectory optimizer that first converts the gradientbased multi-agent trajectory optimization problem into a set of simple matrix-matrix products and then leverages parallel computations over GPUs to accelerate these computations. We demonstrate through a large number of qualitative and quantitative experiments in simulation that our distributed algorithm outperforms existing sequential trajectory optimizers or sampling-based methods for multi-robot planning. Next, we design a collision-aware path planner for robot manipulators operating in the high-dimensional joint space and demonstrate its application across scenes with different types and numbers of obstacles. We finally couple this stochastic optimization-based path planner with a low-level motion planner for the task of pushing objects on a table using a robot manipulator. We also make our software for multi-agent path planning public for open-source development so that our algorithm can be used for further research in this domain

Abstract

The issue of ensuring safety for aerial robots when navigating near obstacles remains a significant concern. There are three distinct aspects to addressing these safety concerns: the controller, which must provide assurances of safety while following a prescribed trajectory; the trajectory planner, which must generate a trajectory that avoids all obstacles; and the perception module, which must accurately localize the robot and obstacles with minimal deviation from actual ground truth. This research study focuses specifically on the first aspect, which pertains to the safe control of aerial robots when maneuvering near static obstacles. The study considers two types of aerial robots, namely an aerial manipulator and a quadrotor unmanned aerial vehicle. The AM considered here is a 2-DOF (degrees-of-freedom) manipulator rigidly attached to a UAV. Our proposed controller structure follows the conventional inner loop PID control for attitude dynamics and an outer loop controller for tracking a reference trajectory. The outer loop control is based on the Model Predictive Control (MPC) with constraints derived using the Barrier Lyapunov Function (BLF) for the safe operation of the AM. BLF-based constraints are proposed for two objectives, viz. 1) To avoid the AM from colliding with static obstacles like a rectangular wall, and 2) To maintain the end effector of the manipulator within the desired workspace. The proposed BLF ensures that the above-mentioned objectives are satisfied even in the presence of unknown bounded disturbances. The capabilities of the proposed controller are demonstrated through high-fidelity non-linear simulations with parameters derived from a real laboratory scale AM. We compare the performance of our controller with other state-of-the-art MPC controllers for AM. To address the problem of flying a UAV inside a tunnel, we propose the implementation of a Model Predictive Control (MPC) framework with constraints based on Control Barrier Function (CBF). The thesis approaches the issue in two distinct ways; first, by maintaining a safe distance from the tunnel walls to avoid the effects of both the walls and ceiling, and second, by minimizing the distance from the walls to effectively manage the nonlinear forces associated with close proximity tasks. Finally, the paper demonstrates the effectiveness of its approach through testing on simulation for various close proximity trajectories with the realistic model of aerodynamic disturbances due to the proximity of the ceiling and boundary walls.

Abstract

Pre-training language models like BERT and then fine-tuning them on downstream tasks has demonstrated state-of-the-art results in various natural language processing tasks. However, pre-training is usually independent of the downstream task, and previous works have shown that this pre-training alone might not be sufficient to capture the task-specific nuances. We propose a novel way to tailor a pretrained BERT model for the downstream task via task-specific masking prior to the standard supervised fine-tuning. Our approach involves first creating a word list specific to the task at hand. For example, if the task is sentiment analysis, we gather a small set of words that represent positive and negative sentiments. If the task is hate speech detection, then we gather a small set of words that represent hate words. If the task is humor detection, then we gather a small set of words that represent humorous words. Next, we use word embeddings to measure the importance of each word in the task using the word list, which we call the word’s task score. Based on the task score, we assign a probability of masking to each word. This probability reflects the likelihood of masking the word during the fine-tuning process. We experiment with different masking functions, including the step function, the linear function, and the exponential function, to determine the best approach for assigning the probability of masking based on the word’s task score. We then use this selective masking strategy to train the BERT model on the masked language modeling (MLM) objective. During MLM training, rather than randomly masking 20% of the input tokens, we selectively mask these input tokens based on their assigned probability of masking, calculated using the word’s task score. Finally, we fine-tune the BERT model on different downstream binary and multi-class classification tasks, such as sentiment analysis, hate speech detection, formality style detection, named entity recognition, and humor detection. Our experiments show that our selective masking strategy outperforms random masking, indicating its effectiveness in fine-tuning the pre-trained BERT model for specific tasks. Overall, our approach provides a more targeted and effective way of fine-tuning pre-trained language models for specific tasks, by incorporating task-specific knowledge between the pre-training and the fine-tuning stages. By selectively masking input tokens based on their importance, we are able to better capture the nuances of a particular task, leading to improved performance on various downstream classification tasks

Abstract

Linear regression is a widely used technique to fit linear models and finds widespread applications across different areas of natural sciences, as well as field such as machine learning and statistics. In most real-world scenarios, however, linear regression problems are often ill-posed or the underlying model suffers from overfitting, leading to erroneous or trivial solutions. This is often dealt with by adding extra constraints, known as regularization. For instance, suppose that we are given N data points {(ai , bi)} N i=1 where ∀i : ai ∈ R d , ∀i : bi ∈ R. The assumption is that there is a vector x such that bi = x T ai +ei , where ei is a random variable (noise) with mean 0. Suppose A is the data matrix of dimension N × d, such that its i th row is the vector ai and b ∈ R N such that b = (b1, · · · , bN ) T . Adding an ℓ2-regularization, the objective is to obtain x that minimizes ∥Ax − b∥ 2 2 + λ∥Lx∥ 2 2 (0.0.1) where L is an appropriately chosen penalty matrix (or regularization matrix) of dimension N ×d and λ > 0 is the regularization parameter, an appropriately chosen constant. This regularization technique is known as general ℓ2 -regularization in the literature. It is a generalization of the well known ridge regression which corresponds to the case when L is the identity matrix. The closed-form solution of the general ℓ2-regularized ordinary least squares problem is given by x = A T A + λLTL
−1 A T b. (0.0.2) A straightforward observation is that even when AT A is singular, regularization ensures that the condition number (ratio of the maximum and the minimum singular values) of the resulting matrix is finite and therefore AT A + λLTL is invertible. Another observation is that when the condition number of A is arbitrarily large (the minimum singular value of A is arbitrarily small), the condition number of the matrix to be inverted in Equation 0.0.2 can be tamed significantly. Since the complexity of most quantum algorithms for the quantum versions of these problems depend the condition number, the complexity reduces significantly. These models can be generalized to situations where the data points are weighted or correlated which incorporate more realistic scenarios. Quantum linear systems solvers are possibly among the most studied quantum algorithms. In this work we develop the first quantum algorithms for ordinary, weighted and generalized least squares with generalized ℓ2- regularization. We assume access to (approximate) block-encodings of the relevant matrices and develop robust versions of quantum singular value transformations to implement linear transformations of these matrices. Our goal is to output a quantum state that is δ-close to |x⟩, the state encoding the solution to the underlying regularized least squares problem. Owing to the generality of the block-encoding framework, our algorithms are applicable to a variety of input models. In order to obtain our results, we work with approximate block-encodings of matrices and apply robust quantum singular value transformations (QSVT) to them. For most quantum linear algebra applications using QSVT, access to perfect block-encoding is assumed, and the robustness of such algorithms is left out. We develop a space-efficient variable time amplitude amplification (VTAA) procedure, which we then use in a variable-time matrix inversion algorithm. A crucial requirement for variable time matrix inversion algorithm is the application of the inversion procedure to the portion of the input state that is spanned by singular values larger than a certain threshold. In order to achieve this, prior results have made use of Gapped Phase Estimation (GPE), which is a simple variant of the standard phase estimation procedure. However, GPE requires extra registers that store the estimates of the phases, which are never used in the variable-time algorithm. In this work, instead of GPE, we make use of a robust version of quantum eigenvalue discrimination (QEVD) using QSVT, which decides whether the eigenvalue of a matrix is above or below a certain threshold without storing the eigenvalue estimates. Given the block-encoding of a matrix A with condition number κ, this procedure for implementing A+ reduces the number of additional qubits required for our variable time by a factor of O(log2 (κ/δ)) as compared to prior results, where δ is the desired accuracy. We show that in order to implement A+, the precision required in the block-encoding of A is determined in turn by the precision in the QEVD procedure. Consequently, this improves both variable-time amplitude amplification and quantum linear systems algorithm: we achieve optimal complexity (linear dependence in the condition number and polylogarithmic dependence on inverse-accuracy) using far fewer ancilla qubits. We then use this to develop our algorithms for ordinary, weighted and generali