Research Hub

    Papers with a full citation graph - what each paper draws on, and who cites it.

    All (51)robotics (6)deep learning (5)backpropagation (3)deep-learning (3)HRI (3)survey (3)algorithms (3)reasoning (2)actuators (2)machine-learning (2)optimization (2)LSTM (2)deep reinforcement learning (2)biped robot (2)representation-learning (2)force control (2)ensemble (2)regression (2)hypothesis-testing (2)statistics (2)autonomous agents (1)web environment (1)evaluation (1)UI automation (1)ReAct (1)function calling (1)agent (1)chain-of-thought (1)prompting (1)agents (1)sensors (1)review (1)automatic-differentiation (1)calculus (1)pose estimation (1)openpose (1)part affinity fields (1)computer vision (1)skeletal tracking (1)OCR (1)text recognition (1)CRNN (1)CTC (1)locomotion (1)Sim-to-Real (1)Transformer (1)attention (1)NLP (1)machine translation (1)object detection (1)YOLO (1)real-time vision (1)bounding boxes (1)gradient-descent (1)adam-optimizer (1)DQN (1)Q-learning (1)Atari games (1)transfer learning (1)feature extraction (1)CNNs (1)generalization (1)vector-spaces (1)word-embeddings (1)cosine-similarity (1)nlp (1)CNN (1)ImageNet (1)image classification (1)IMU (1)orientation filter (1)quaternion (1)ROS (1)Robot Operating System (1)open-source (1)middleware (1)service robot (1)elderly care (1)robot arm (1)imitation learning (1)learning from demonstration (1)recommender systems (1)matrix factorization (1)SVD (1)collaborative filtering (1)Netflix Prize (1)safety (1)collaborative robots (1)standards (1)robotic hand (1)pneumatic actuators (1)force sensor (1)design (1)ZMP (1)balance (1)control (1)human-robot interaction (1)social robots (1)service robotics (1)Random Forest (1)decision tree (1)classification (1)biped walking (1)inverted pendulum (1)gait pattern (1)humanoid robot (1)random forest (1)decision trees (1)supervised learning (1)motion planning (1)RRT (1)random tree (1)impedance (1)hybrid control (1)search engines (1)PageRank (1)eigenvectors (1)graph theory (1)Markov chains (1)localization (1)particle filter (1)navigation (1)mobile robot (1)mechanical design (1)ASIMO (1)humanoid (1)RNN (1)sequence modeling (1)vanishing gradient (1)visual servoing (1)image-based (1)position-based (1)lasso (1)regularization (1)feature selection (1)cross-validation (1)model evaluation (1)bootstrap (1)model selection (1)reinforcement learning (1)temporal difference (1)prediction (1)TD (1)edge detection (1)image processing (1)Canny filter (1)hysteresis (1)neural networks (1)representations (1)neural-networks (1)credit-assignment (1)convex-optimization (1)linear-programming (1)interior-point-methods (1)computational-complexity (1)least squares (1)history of statistics (1)Gauss (1)matrix multiplication (1)computational complexity (1)numerical linear algebra (1)K-Means (1)clustering (1)unsupervised learning (1)data analysis (1)logistic regression (1)binary data (1)maximum likelihood (1)kinematics (1)Denavit-Hartenberg (1)transformation matrix (1)markov-chain (1)monte-carlo (1)mcmc (1)statistical-mechanics (1)PID (1)Ziegler-Nichols (1)tuning (1)industrial control (1)statistical-significance (1)neyman-pearson (1)decision-theory (1)t-distribution (1)sample-size (1)bayes-theorem (1)probability (1)history-of-mathematics (1)

    51 papers

    2023arXiv preprint arXiv:2307.13854, presented at NeurIPS 2023

    WebArena: A Realistic Web Environment for Building Autonomous Agents

    Finding: Results show that LLM-based agents (like GPT-4) can successfully complete about 40% of tasks, which is a significant achievement, but still far from human performance (~90%).

    autonomous agentsweb environment
    2022arXiv:2210.03629, DOI: 10.48550/arXiv.2210.03629

    ReAct: Synergizing Reasoning and Acting in Language Models

    Finding: ReAct outperformed prior methods on interactive decision-making tasks and improved human interpretability and trustworthiness.

    ReActfunction calling
    2022Advances in Neural Information Processing Systems (NeurIPS) 2022, arXiv:2201.11903

    Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

    Finding: This method significantly improved accuracy on math, logical reasoning, and complex QA tasks, even for smaller models.

    chain-of-thoughtreasoning
    2018IEEE Sensors Journal, vol. 18, no. 16, pp. 6484-6495, DOI: 10.1109/JSEN.2018.2846743

    A Review of Sensors and Actuators for Robotics

    Finding: MEMS-based sensors and electric actuators (e.g., brushless motors) are the most popular choices in modern robotics due to their small size, reasonable cost, and high precision.

    sensorsactuators
    2017Journal of Machine Learning Research (JMLR), Vol. 18, pp. 1-43 — https://jmlr.org/papers/v18/17-468.html

    Automatic Differentiation in Machine Learning: a Survey

    Finding: Reverse-mode AD is the foundational algorithm that powers the training of virtually all deep learning models — commonly known in the ML community as backpropagation.

    automatic-differentiationbackpropagation
    2017IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (DOI: 10.1109/CVPR.2017.143)

    Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

    Finding: This bottom-up architecture achieved high accuracy while maintaining real-time processing speeds, crucially remaining fast regardless of how many people or objects were in the image.

    pose estimationopenpose
    2017IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 11, pp. 2298-2304, DOI: 10.1109/TPAMI.2016.2646691

    An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition

    Finding: The proposed method achieved competitive and superior accuracy on standard benchmarks like IIIT-5K and ICDAR, while being end-to-end trainable and requiring no manual segmentation.

    OCRtext recognition
    2017arXiv preprint arXiv:1707.02286 (presented at NeurIPS 2017)

    Emergence of Locomotion Behaviours in Rich Environments

    Finding: The trained policies exhibited emergent behaviors like walking, running, jumping, and even leaping across various environments, demonstrating robust locomotion.

    locomotiondeep reinforcement learning
    2017Advances in Neural Information Processing Systems (NeurIPS) 2017, pp. 5998-6008, arXiv:1706.03762

    Attention Is All You Need

    Finding: The Transformer achieved state-of-the-art translation quality while being significantly faster to train due to parallelization.

    Transformerattention
    2016IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (DOI: 10.1109/CVPR.2016.91)

    You Only Look Once: Unified, Real-Time Object Detection

    Finding: YOLO achieved state-of-the-art performance at incredibly high speeds (up to 45 frames per second on standard hardware, and 155 fps for smaller versions), proving that a single end-to-end network could achieve both high accuracy and real-time processing.

    object detectionYOLO
    2015International Conference on Learning Representations (ICLR 2015) — arXiv:1412.6980, https://arxiv.org/abs/1412.6980

    Adam: A Method for Stochastic Optimization

    Finding: Adam combines the advantages of AdaGrad (handles sparse gradients) and RMSProp (works well in non-stationary settings), and performs well empirically across problems.

    optimizationgradient-descent
    2015Nature, vol. 518, no. 7540, pp. 529-533, DOI: 10.1038/nature14236

    Human-level control through deep reinforcement learning

    Finding: DQN achieved human-level or better performance on 49 Atari games, marking the first major success of Deep RL.

    DQNdeep reinforcement learning
    2014Advances in Neural Information Processing Systems (NeurIPS) 2014, pp. 3320-3328

    How transferable are features in deep neural networks?

    Finding: They found that features become increasingly specific to the original task in higher layers, but still transfer well to related tasks. Transferability degrades as the divergence between tasks increases.

    transfer learningfeature extraction
    2013ICLR 2013 Workshop Track; arXiv preprint arXiv:1301.3781 — https://arxiv.org/abs/1301.3781

    Efficient Estimation of Word Representations in Vector Space

    Finding: The models learned high-quality word vectors from 1.6 billion words in under a day, orders of magnitude faster than earlier neural approaches. Strikingly, the resulting vector space encoded semantic and syntactic relationships as directions: vector arithmetic such as king − man + woman produced a vector closest to queen, showing that regularities in language were captured as consistent geometric offsets.

    vector-spacesword-embeddings
    2012Advances in Neural Information Processing Systems (NeurIPS) 2012, pp. 1097-1105

    ImageNet Classification with Deep Convolutional Neural Networks

    Finding: AlexNet won the ImageNet 2012 competition by a large margin (15.3% top-5 error vs 26.2% for the second best), proving that deep CNNs are highly effective on large-scale data.

    CNNImageNet
    2010Technical Report, University of Bristol, DOI: 10.1007/978-3-642-21799-3_1 (also published in IEEE Sensors Journal)

    An efficient orientation filter for inertial and inertial/magnetic sensor arrays

    Finding: The Madgwick filter can operate at high sampling rates (>500 Hz) with low error, achieving performance comparable to extended Kalman filters in practice.

    IMUorientation filter
    2009Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) Workshop on Open Source Software, pp. 1-6

    ROS: an open-source Robot Operating System

    Finding: ROS quickly became the industry standard for robot development, accelerating robotics research due to its open-source nature and active community.

    ROSRobot Operating System
    2009Proceedings of the 2009 IEEE International Symposium on Robot and Human Interactive Communication (RO‑MAN), pp. 331-336, DOI: 10.1109/ROMAN.2009.5326349

    The Care‑O‑bot 3 – A Service Robot for Home Environments

    Finding: Care‑O‑bot 3 was shown to effectively perform tasks like object handling, drink serving, and video communication, receiving positive feedback in user trials.

    service robotelderly care
    2009Robotics and Autonomous Systems, vol. 57, no. 5, pp. 469-483, DOI: 10.1016/j.robot.2008.10.024

    A survey of robot learning from demonstration

    Finding: They show that learning from demonstration can effectively transfer complex skills to robots, and combining it with RL can yield results surpassing the expert.

    imitation learningrobotics
    2009IEEE Computer (DOI: 10.1109/MC.2009.263)

    Matrix Factorization Techniques for Recommender Systems

    Finding: Matrix factorization yielded superior predictive accuracy compared to nearest-neighbor techniques. By modeling implicit feedback, temporal effects, and confidence levels within the factorization framework, the team successfully won the $1 million Netflix Prize.

    recommender systemsmatrix factorization
    2008Robotics and Autonomous Systems, vol. 56, no. 3, pp. 241-252, DOI: 10.1016/j.robot.2007.08.005

    Safety in human-robot interaction: a survey

    Finding: They show that combining force/torque sensors with adaptive control can significantly reduce risks, and existing standards (e.g., ISO) provide a solid framework.

    safetycollaborative robots
    2006IEEE Robotics & Automation Magazine, vol. 13, no. 3, pp. 38-46, DOI: 10.1109/MRA.2006.1668325

    The Shadow Dexterous Hand: A High-Fidelity Humanoid Hand for Telemanipulation

    Finding: The hand was shown to perform delicate tasks like grasping an egg or playing piano, demonstrating its capability as a human assistant in hazardous environments.

    robotic handpneumatic actuators
    2004International Journal of Humanoid Robotics, vol. 1, no. 1, pp. 157-173, DOI: 10.1142/S0219843604000083

    Zero-moment point – thirty five years of its life

    Finding: ZMP has become a cornerstone of bipedal stability control, and many successful algorithms are based on it.

    ZMPbalance
    2003Robotics and Autonomous Systems, Vol. 42, Issues 3–4, pp. 143–166, DOI: 10.1016/S0921-8890(02)00372-X

    A survey of socially interactive robots

    Finding: A unified framework for understanding socially interactive robots and identification of open research issues in HRI.

    human-robot interactionsocial robots
    2001Machine Learning, vol. 45, no. 1, pp. 5-32, DOI: 10.1023/A:1010933404324

    Random Forests

    Finding: Random Forest was shown to outperform single trees and other ensemble methods, handling high-dimensional and noisy data well.

    Random Forestdecision tree
    2001Proceedings of the 2001 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), vol. 1, pp. 239-246, DOI: 10.1109/IROS.2001.973315

    The 3D Linear Inverted Pendulum Mode: A simple modeling for a biped walking pattern generation

    Finding: The LIPM model was shown to predict walking motion well for biped robots, and controllers based on it provide stable locomotion.

    biped walkinginverted pendulum
    2001Machine Learning, Vol. 45, pp. 5–32, DOI: 10.1023/A:1010933404324

    Random Forests

    Finding: Generalisation error converges as the number of trees grows; the method is robust to noise and also yields variable-importance measures.

    random forestdecision trees
    2000Proceedings of the 2000 IEEE International Conference on Robotics and Automation (ICRA), vol. 2, pp. 995-1001, DOI: 10.1109/ROBOT.2000.844730

    RRT-Connect: An Efficient Approach to Single-Query Path Planning

    Finding: RRT-Connect significantly reduces planning time compared to standard RRT, making it highly effective for single-query problems.

    motion planningRRT
    2000Industrial Robot: An International Journal, vol. 27, no. 4, pp. 278-285, DOI: 10.1108/01439910010334289

    Force control in robotic manipulation: a review

    Finding: Impedance control is better suited for soft environments, while hybrid control is preferred for rigid surfaces; the choice depends on the application.

    force controlimpedance
    1999Stanford InfoLab Technical Report (http://ilpubs.stanford.edu:8090/422/)

    The PageRank Citation Ranking: Bringing Order to the Web

    Finding: The values in this principal eigenvector represented the steady-state probability of a random web surfer landing on any given page. This metric, 'PageRank', provided an incredibly robust and highly relevant ranking of web pages.

    search enginesPageRank
    1999Proceedings of the 1999 IEEE International Conference on Robotics and Automation (ICRA), pp. 1322-1328, DOI: 10.1109/ROBOT.1999.772544

    Monte Carlo Localization for Mobile Robots

    Finding: The particle filter can handle multi‑modal distributions and outperforms the Kalman filter in real‑world environments with sensor noise.

    localizationparticle filter
    1998Proceedings of the 1998 IEEE International Conference on Robotics and Automation (ICRA), vol. 2, pp. 1321-1326, DOI: 10.1109/ROBOT.1998.677262

    The development of Honda humanoid robot

    Finding: ASIMO achieved stable walking, stair climbing, and even running, demonstrating that careful mechanical design combined with proper control can produce human-like performance.

    mechanical designASIMO
    1997Neural Computation, vol. 9, no. 8, pp. 1735-1780, DOI: 10.1162/neco.1997.9.8.1735

    Long Short-Term Memory

    Finding: LSTM outperformed vanilla RNNs on tasks like speech recognition, machine translation, and text generation, successfully modeling long-term dependencies.

    RNNLSTM
    1996IEEE Transactions on Robotics and Automation, vol. 12, no. 5, pp. 651-670, DOI: 10.1109/70.538972

    A tutorial on visual servo control

    Finding: Image-based servoing is generally faster and more robust, but may suffer from field-of-view limitations, while position-based servoing requires accurate 3D estimation but facilitates path planning.

    visual servoingimage-based
    1996Journal of the Royal Statistical Society, Series B, vol. 58, no. 1, pp. 267-288, DOI: 10.1111/j.2517-6161.1996.tb02080.x

    Regression Shrinkage and Selection via the Lasso

    Finding: Lasso was shown to produce more interpretable models compared to Ridge regression and works well in high-dimensional settings.

    lassoregularization
    1995Proceedings of the 14th International Joint Conference on Artificial Intelligence (IJCAI), vol. 2, pp. 1137-1143

    A study of cross-validation and bootstrap for accuracy estimation and model selection

    Finding: 10-fold cross-validation generally offers a good bias-variance trade-off and is often the best choice.

    cross-validationmodel evaluation
    1988Machine Learning, vol. 3, no. 1, pp. 9-44, DOI: 10.1007/BF00115009

    Learning to predict by the methods of temporal differences

    Finding: TD methods were shown to converge faster than Monte Carlo methods in prediction problems and became a cornerstone of modern RL.

    reinforcement learningtemporal difference
    1986IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-8, no. 6, pp. 679-698, DOI: 10.1109/TPAMI.1986.4767851

    A Computational Approach to Edge Detection

    Finding: The Canny edge detector was shown to be statistically optimal and outperformed previous methods, becoming the gold standard for edge detection.

    edge detectionimage processing
    1986Nature, vol. 323, no. 6088, pp. 533-536, DOI: 10.1038/323533a0

    Learning representations by back-propagating errors

    Finding: Backpropagation effectively trains multi-layer networks for nonlinear tasks like pattern recognition and representation learning.

    neural networksbackpropagation
    1986Nature, Vol. 323, pp. 533-536 — https://doi.org/10.1038/323533a0

    Learning representations by back-propagating errors

    Finding: Hidden units learn to represent important task features on their own (e.g. 'person', 'generation', 'nationality' in a family-tree problem) — deep networks learn abstract representations, not just memorize.

    backpropagationneural-networks
    1984Combinatorica, Vol. 4, No. 4, pp. 373-395 — https://doi.org/10.1007/BF02579150

    A new polynomial-time algorithm for linear programming

    Finding: Proved the algorithm runs in polynomial time, answering the open question, and claimed practical speed advantages over Simplex for large problems.

    convex-optimizationlinear-programming
    1981The Annals of Statistics, vol. 9, no. 3, pp. 465-474, DOI: 10.1214/aos/1176345451

    Gauss and the invention of least squares

    Finding: Gauss introduced the least squares method in 1809 to solve problems in astronomy and geodesy, and it remains one of the most important statistical methods.

    least squaresregression
    1969Numerische Mathematik (DOI: 10.1007/BF02165411)

    Gaussian Elimination is not Optimal

    Finding: By recursively applying this 7-multiplication technique, Strassen proved that matrix multiplication could be performed in O(N^2.807) time, fundamentally breaking the assumed O(N^3) barrier and spawning a new field of algebraic complexity theory.

    matrix multiplicationcomputational complexity
    1967Proceedings of the 5th Berkeley Symposium on Mathematical Statistics and Probability, vol. 1, pp. 281-297

    Some methods for classification and analysis of multivariate observations

    Finding: K-Means was shown to be a simple and effective clustering method, though it may converge to local optima.

    K-Meansclustering
    1958Journal of the Royal Statistical Society, Series B, vol. 20, no. 2, pp. 215-242, DOI: 10.1111/j.2517-6161.1958.tb00292.x

    The regression analysis of binary sequences (with discussion)

    Finding: Logistic regression was shown to be an effective tool for modeling the probability of a binary event based on predictor variables.

    logistic regressionbinary data
    1955Journal of Applied Mechanics, vol. 22, no. 2, pp. 215-221

    A kinematic notation for lower-pair mechanisms based on matrices

    Finding: The DH method became a worldwide standard for modeling robot arm kinematics and is used in all industrial and research robots.

    kinematicsDenavit-Hartenberg
    1953The Journal of Chemical Physics (https://doi.org/10.1063/1.1699114)

    Equation of State Calculations by Fast Computing Machines

    Finding: Demonstrated that sampling from a constructed Markov Chain allows accurate simulation of physical systems and computation of thermodynamic properties.

    markov-chainmonte-carlo
    1942Transactions of the American Society of Mechanical Engineers, vol. 64, no. 8, pp. 759-768

    Optimum Settings for Automatic Controllers

    Finding: The Ziegler-Nichols methods quickly became industry standard for PID tuning and are still used as a good starting point.

    PIDZiegler-Nichols
    1933Philosophical Transactions of the Royal Society of London. Series A (https://doi.org/10.1098/rsta.1933.0009)

    On the Problem of the Most Efficient Tests of Statistical Hypotheses

    Finding: Proved that the likelihood ratio test is the most powerful test for comparing two simple hypotheses at a fixed significance level.

    hypothesis-testingstatistical-significance
    1908Biometrika (https://doi.org/10.1093/biomet/6.1.1)

    The Probable Error of a Mean

    Finding: Proved that for small samples, the distribution of the mean has heavier tails than the normal distribution, and provided tables of critical values for statistical significance.

    t-distributionstatistics
    1763Philosophical Transactions of the Royal Society of London (https://doi.org/10.1098/rstl.1763.0053)

    An Essay towards Solving a Problem in the Doctrine of Chances

    Finding: The derivation of the basic formulation of posterior distributions, laying the foundational framework for what is now known as Bayesian inference.

    bayes-theoremprobability