Predicting and Protecting: Safe Robot Collaboration in a Human World
A new framework dynamically adapts robot behavior to ensure safe and reliable interaction with humans, accounting for the inherent unpredictability of human actions.
Science
The AI-Powered Software Sprint
Artificial intelligence is rapidly reshaping software engineering, promising faster development cycles and more innovative applications.
Decoding Movement: New Insights into Upper-Limb Rehabilitation
A new study leverages advanced data analysis to better understand how patients recover arm function after stroke, paving the way for more effective robotic therapies.
Mining the Omics Landscape with AI
New research explores how artificial intelligence can automatically unlock valuable insights hidden within the growing flood of omics data.
Beyond Consumption: How Reading Fuels Creativity
New research proposes a framework for understanding reading not as passive intake, but as an active creative process, with implications for how we design digital reading tools.
Beyond the Black Box: Defining AI’s Core Components
A new framework clarifies the often-blurred line between AI models and the larger systems they power, paving the way for more effective oversight.
Growing Robots: A New Era of Adaptive Hardware
![Inspired by the vascular system of the <i>Catocala fraxini</i> moth, a novel system generates receptors on demand within an artificial circulatory network, enabling responsive behavior such as controlling wing-like flapping or visual signaling through localized physical updates and closed-loop control-mimicking the hemolymph-carrying veins observed in the moth’s wings at a scale of [latex]100 \, \mathrm{\SIUnitSymbolMicro m}[/latex].](https://arxiv.org/html/2603.09473v1/x1.png)
Researchers have demonstrated a method for building robotic capabilities on demand by vascularizing a robotic system and using it to ‘grow’ new components in situ.
Building AI That Explains Itself: A New Approach to Innovation
Researchers have developed a system that moves beyond simply generating answers, focusing instead on creating an AI capable of tracing its reasoning and verifying its conclusions.
Safeguarding Robot Motion in Uncertain Worlds
![The system achieves risk-bounded motion planning for robotic manipulators operating under uncertainty by integrating a neural network - predicting states [latex]X_{N,H}[/latex] and costs [latex]C_{N,H}[/latex] - with Model Predictive Path Integration, refining control inputs [latex]u^<i>[/latex], and a supervisory logic enforcing safety through simulation-based collision risk assessment and contact force computation, ultimately guiding stochastic optimization with a novel cost function [latex]\hat{c}(x_t, u^</i>)[/latex] that ensures adherence to predefined risk constraints within a perception-action loop utilizing noisy state estimations [latex]x_t[/latex].](https://arxiv.org/html/2603.09083v1/x1.png)
Researchers have developed a new approach to generating safe trajectories for robotic arms operating in complex and unpredictable environments.
Beyond Logic: Teaching AI to Understand People
New research introduces a framework for building language models that reason about social situations more like humans do, moving past purely logical deduction.