🚀 Robotics Breakthroughs

This paper presents the first experimental demonstration of non-line-of-sight (NLOS) imaging applied to autonomous navigation, using single-photon LiDAR to detect hidden objects. It enables robots to perceive and avoid obstacles around corners, significantly expanding perception capabilities beyond line-of-sight.

NVIDIA Research introduced NeRD, Dexplore, and VT-Refine at CoRL 2025, enhancing simulation with learned dynamics, enabling efficient exploration, and combining vision-tactile sensing for bimanual assembly. These innovations represent a first-of-its-kind capability in integrating multiple neural approaches for practical robot learning.

Figure AI's F.03 humanoid robot autonomously sorted nearly 250,000 packages over 200 hours in a warehouse stress test, achieving near-human speed. This marks a significant milestone in deploying general-purpose humanoid robots for real-world logistics automation, demonstrating reliability and scalability.

McGill engineers created ultra-thin graphene oxide sheets that fold like origami, enabling soft robots to walk, twist, and sense motion. This addresses the key scalability challenge in soft robotics by combining actuation and sensing in a single material.

NVIDIA Research's Eureka AI agent autonomously generates reward algorithms for robot training, outperforming human-written algorithms on over 80% of complex tasks. This represents a significant advancement in autonomous skill acquisition for robotics, potentially accelerating the development of general-purpose robots.

A humanoid robot completed a half-marathon faster than any human, demonstrating unprecedented endurance and locomotion control. This milestone suggests rapid progress in bipedal robotics and could accelerate adoption in logistics and emergency response.

Aquila Earth achieved a world-first by wirelessly powering a moving warehouse robot for 24 hours using infrared laser beams, setting records for highest total laser power transferred and longest continuous duration. This demonstration validates power-beaming as a viable solution for continuous autonomous operations, potentially eliminating battery downtime in logistics and other industries.

This paper presents a novel actuator design using three parallel motors to independently control stiffness and damping, enabling safe and efficient human-robot collaboration in dynamic tasks. It addresses a key limitation of traditional rigid actuators by providing physical compliance adjustability.

RoboParty's Roboto Origin is the first fully open-source bipedal humanoid robot, providing developers with complete hardware and software access. This democratizes humanoid robotics development, potentially accelerating innovation and lowering barriers to entry.

MIT Technology Review highlights a breakthrough in AI that enables robots to learn complex tasks rapidly by merging diverse data sources. This advancement significantly accelerates robot training for warehouse and home applications, marking a notable step toward versatile autonomous systems.

Argonne National Laboratory's 2025 breakthroughs include Polybot, an automated materials lab integrating robotics and machine learning, and AI-enhanced catalyst research that combines robotics with world-class facilities. These advances represent a first-of-its-kind capability in autonomous materials discovery and catalyst optimization.

Figure AI's humanoid robots autonomously processed nearly 250,000 packages over 200 continuous hours without any hardware failures, demonstrating a new level of sustained autonomy in logistics. This marks a significant milestone in humanoid robot reliability and operational endurance for real-world industrial applications.

This advancement transforms LiDAR from a simple rangefinder into a full-color 3D camera that operates independently of external cameras in all weather conditions. The 6D capability adds color and environmental awareness, potentially revolutionizing autonomous driving perception systems.

This paper presents a novel LiDAR design that mimics the retina's non-uniform angular resolution, drastically reducing cost while maintaining high performance for autonomous robots. The approach uses incommensurable scanning to achieve a dense central field and sparse periphery, improving object detection efficiency. It represents a significant step toward accessible, high-performance sensing for robotics.

MIT's EES algorithm allows robots to self-assess and refine skills in unfamiliar environments with far fewer data samples than traditional reinforcement learning. This represents a significant advance in sample efficiency for autonomous learning, potentially accelerating deployment of adaptive robots in real-world settings.

Purdue University students built Purdubik's Cube, a robot that solves a puzzle cube in 0.103 seconds, nearly three times faster than the previous record. This demonstrates exceptional engineering in speed, precision, and control systems.

This paper presents a MEMS mirror LiDAR system that allows independent control of sensor field-of-view, decoupling robot motion from perception. This innovation enables micro-robots and UAVs to maintain stable sensing regardless of platform movement, a first-of-its-kind capability for lightweight platforms.

This research demonstrates that humans can learn to control a supernumerary robotic limb and generalize those motor skills to new tasks, showing that the brain can adapt to an extra limb beyond the biological body. The findings have implications for assistive technology and human augmentation.

This research presents a highly deformable membrane with integrated proprioceptive sensing that can reconstruct its 3D shape in real time. This enables soft robots to perceive their own deformation without external cameras, a key step toward autonomous manipulation in unstructured environments.

Humanoid's Alpha bipedal robot achieved stable walking just 48 hours after final assembly, a process that typically takes weeks or months. This milestone, reached after only five months from initial design to working prototype, represents a significant acceleration in bipedal robot development.

Researchers at Harvard have created a new liquid crystal elastomer material that can deform in opposite directions when heated, enabling unprecedented control in soft robotics. This peer-reviewed advance offers a first-of-its-kind capability for applications in wearable devices and smart textiles.

NVIDIA's Isaac GR00T N1 is the first open, customizable foundation model for humanoid robot reasoning and skills, combining simulation frameworks and an open-source physics engine developed with Google DeepMind and Disney Research. This marks a significant milestone by democratizing access to advanced AI for humanoid robotics, potentially accelerating development across the industry.

Google DeepMind introduces Gemini Robotics, a family of vision-language-action models that enable humanoid robots to perceive, reason, and perform complex multi-step tasks with generalization across different robot embodiments. This represents a significant step toward general-purpose robotics, integrating advanced AI reasoning with physical action.

This paper presents a digital twin framework for automating chemistry laboratories using robotics, published in Nature Computational Science. It enables real-time simulation and control of robotic systems for chemical synthesis and analysis, representing a first-of-its-kind integration of digital twins with lab automation. The work is peer-reviewed and demonstrates a novel capability that could accelerate chemical research.