All papers

The authors introduce a training loop where multimodal models use text and visual abilities alongside program execution to generate their own training data.

The paper surveys artificial intelligence techniques for detecting and mitigating jamming and interference in 5G and 6G wireless communication networks.

The paper investigates the gap between raw obfuscated email source and recipient-visible text, evaluating lightweight classifiers under different obfuscations.

This paper introduces an adaptive framework to detect radio jamming attacks within machine learning integrated 5G Open RAN architectures.

This paper uses Bayesian inference combined with machine learning models to detect and classify jamming attacks in 5G networks.

This paper presents a convolutional neural network model to detect and classify facial acne across four areas of the face.

This paper implements a lightweight binary neural network on a RISC-V processor to detect attacks on vehicle CAN bus networks.

This paper presents a dataset, benchmark, and coarse-to-fine learning scheme to detect tiny oriented objects in imagery.

This paper presents a framework and a dataset to generate holistic full-body co-speech gestures from audio and masked movement priors.

The authors present a lightweight convolutional network architecture for classifying multiple ophthalmic diseases from imbalanced retinal fundus images.

This paper presents an optimization method to separate crossing white matter fiber orientations in diffusion magnetic resonance imaging.

The authors propose an interpretable histopathology classification architecture evaluated across multiple hospital sites with verifiable decision records.

The paper develops a joint Bayesian reconstruction method with uncertainty quantification for multi-contrast X-ray computed tomography from sparse data.

This work develops a smooth-to-hard spectral regularization penalty with convergence stationarity guarantees for solving general linear inverse problems.

The paper introduces a framework to learn adaptive metric updates for proximal splitting algorithms while retaining convergence guarantees.

The authors analyze causal internal pathways in vision-language-action models to identify where visual and instruction features affect predicted actions.

This work uses block-sparse featurizers to extract interpretable representations for steering robot foundation model behavior and shaping training data.

The paper presents an absolute vehicle heading estimation framework that optimizes a skylight polarization model using polarimetric camera measurements.

The authors present a physical robotic framework for non-invasive automated testing to detect data loss in Android mobile applications.

This technical note compares two observer-based control schemes on a pendulum system to evaluate chattering attenuation under nonlinear friction.

This paper introduces an uncertainty-aware method to select effective gaits for damaged legged robots during continuous reset-free operation.

The authors present an adaptive mission planning method using ergodic search to maximize information gathering for solar-powered autonomous vessels.

The paper trains a reinforcement learning policy to recover vehicle stability beyond normal handling limits using learned drifting maneuvers.

Formulates a periodic optimal control problem to optimize speed trajectories and balance energy consumption with information gain in mobile robots.

Develops a constructive control barrier function approach using distance bubbles to enforce safe navigation in second-order dynamical systems.

Introduces an augmented Lagrangian trajectory optimization method using gradient flow to compute feasible motions for contact-constrained robots.

Develops an adaptive trajectory tracking controller for uncertain autonomous underwater vehicles operating with saturated thruster inputs.

Analyzes the geometric and transport conditions under which robots can use reversing wind headings to locate odour sources.

Introduces an agentic framework that extracts spatial cues from passive human videos to guide mobile robot manipulation via executable code policies.

Formulates an optimization-based control method to suppress slip-induced falls in remotely operated motorcycles during cornering on slippery surfaces.

Proposes a generative motion and impedance optimization framework combining reinforcement learning priors with model-based refinement for contact-rich tasks.

Presents a data alignment framework that converts human egocentric videos into controller-compatible training demonstrations for humanoid robots.

Formalizes a decision-theoretic framework and measurement standard for scheduling slow module invocations in dual-system vision-language-action models.

Introduces a benchmark evaluating few-shot generalization and knowledge transfer across related task families in dexterous robotic manipulation.

Introduces a hybrid LiDAR-inertial neural SLAM system combining geometric estimation with implicit neural mapping for real-time onboard navigation.

Proposes an agentic framework using procedural assistance memory to guide proactive robots in assisting users during daily routines.

Demonstrates closed-loop autonomous navigation of microbots using real-time object detection and magnetic steering along complex trajectories.

Proposes a model predictive path integral controller that samples execution-time initial states to mitigate real-world robotic planning delays.

Presents a model predictive contouring control framework for robotic manipulators tracking paths under moving reference frames.

Introduces a GPU-accelerated differentiable simulator to optimize contact-rich manipulation trajectories for robot hands via batched model predictive control.

Presents a plug-and-play pedestrian dead reckoning sensor module for wearable navigation in GPS-degraded environments.

Proposes a hierarchical delivery routing framework across building, floor, and room levels that incorporates access and privacy constraints.

Presents a visual safety certification framework using depth-normalized monocular optical flow to certify collision risk without 3D geometric reconstruction.

Introduces a robot manipulation policy that caches vision-language model outputs once per episode to accelerate action prediction without recomputing task context.

Develops a modular inverse-dynamics workflow to recover unique actuator and joint wrenches in closed-chain robotic mechanisms.

Introduces a dataset and spatial-memory policy model to evaluate sound-triggered mobile robot manipulation across unseen multi-scene environments.

Proposes a 2D image feedback perception and control framework for autonomous endovascular navigation using centerline graphs and rule-based controllers.

Presents a wearable tag and cloud platform for tracking workers seamlessly across indoor and outdoor industrial environments.

Presents a conversational editing framework that aligns language instructions with measurable kinematic events for humanoid motion generation.

Proposes a two-level on-policy distillation method to consolidate specialized skills from multiple teacher models into one flow-based policy.

Introduces an attribution-gated estimation framework to separate intended user motion from voluntary torque capability in assistive robotics.

Fuses neural and muscular signals into an adaptive exoskeleton control strategy to distinguish user inability from mental disengagement.

Develops a selective query selection and reward framework to make vision-language models efficient preference guides for robotic manipulation.

Proposes test-time hidden-state steering along expert demonstration geometries to improve the robustness of imitation learning policies.

Introduces a multimodal world model and dataset to anticipate impending execution anomalies during contact-rich robotic manipulation.

Presents a unified policy learning prehensile and non-prehensile actions together for reaching object poses across diverse objects and grippers.

Identifies metric flaws in the Bench2Drive benchmark where comfort scores reward stationary vehicles, and proposes a distance-based correction.

Extracts object trajectories from single-shot human demonstration videos to train generalizable manipulation policies without explicit action labels.

Formulates a time-optimal ergodic search method that accounts for target return phases to reduce total robotic retrieval time.

Introduces an augmented Lagrangian trajectory optimizer based on affine geometric heat flow to handle constrained mechanical systems.