Humanoid robots face a fundamental engineering challenge: the hardware needed for movement competes with space and weight, limiting AI-driven actions. This report explores how power architecture, sensing, and safety systems must evolve to enable practical humanoids.
MIT Lincoln Laboratory demonstrated a magnetic-field communication system through 3.6 feet of Arctic ice using an ROV, achieving 1.2 KB/s data transfer. The test, conducted under extreme conditions, marks a step toward a large-scale underwater sensor network for climate and military monitoring.
At IFA 2026 in Berlin, Paris-based design duo Batten & Kamp presented 'Beauty Seeker', a six-legged AI robot transformed into a mobile vase carrying an orchid. By stripping its advanced features, they provoke reflection on cohabitation, threat, and the value of attention in a tech-saturated world.
Chinese scientists unveiled Jier 2.0, the world's first humanoid embodied intelligent breeding robot, at the 33rd China Beijing Seed Industry Conference. It autonomously pollinates tomatoes in under 10 seconds, integrating gene editing, AI vision, and dual-arm operation.
Researchers from Tohoku University and VISTEC have developed a robot that learns to walk by observing stick insects, mastering diverse terrains in under an hour. The approach uses adversarial inverse reinforcement learning to capture general principles rather than fixed gaits.
Researchers at North Carolina State University developed a soft ring robot that jumps continuously under constant infrared light, using geometric design and photoresponsive materials instead of electronics. This physical intelligence approach enables self-resetting and diverse locomotion modes, offering a new path for miniature soft robots.
Global military research is exploring humanoid robots for combat, but they remain in engineering validation phase. Despite China's supply chain advantages, challenges in reliability, endurance, and cost keep humanoids from deployment, with dual-use conversion driving progress.
SuperMap Software stated that its spatial intelligence base does not directly support 3D data acquisition for embodied robots, clarifying its role in the robotics ecosystem.
SuperMap Software announced that its spatial agent, SuperMap AgentX, provides core spatial intelligence capabilities for robotics and unmanned equipment, enabling spatial perception, analysis, task planning, and multi-machine collaboration.
A Chinese medical team has performed the world's first AI-ultrasound-robot-guided intervention for congenital heart disease, marking a new phase where AI assists in core surgical decisions. This signals a shift in medical robotics from mechanical precision to intelligent, real-time navigation, with potential to improve accuracy and the dissemination of expert care.
Astribot's SmoothRL framework enables robots to learn and improve during real-world execution. Tests show success rates soaring from 39% to 94% in dynamic tasks, marking a shift toward continuous post-deployment improvement.
OpenAI's GPT-6 Astra autonomously completed Valve's puzzle game Portal over 24 hours using 3,336 tool calls, marking a milestone in long-horizon AI decision-making in 3D environments. Separately, bankrupted iRobot returned with a novel dual-robot cleaning system, betting on a niche to compete against dominant Chinese rivals.