Robots Learn to Think in 3D, Work in Groups, and Fix Themselves

Research

SpatialVLA: A Vision-Language-Action Model for 3D Spatial Reasoning

This paper tackles the fundamental problem of robots being bad at understanding 3D space. It introduces a new VLA that explicitly reasons about object geometry and spatial relationships, which is critical for any manipulation task that isn't a fixed pick-and-place.

Self-Reconfigurable Robots: A Review of Mechanisms and Control

Forget simple grippers. This review covers robots that can physically reshape themselves to adapt to tasks or recover from damage. It's the blueprint for building systems that are resilient by design, not just by redundancy.

Analysis

A Survey of Multi-Robot Collaboration: From Theory to Real-World Deployment

This is a comprehensive map of the multi-agent robotics landscape, covering coordination, communication, and task allocation. For builders, it's a cheat sheet for avoiding common pitfalls when scaling from a single bot to a fleet.

Tools

NVIDIA Isaac Sim Update: Enhanced Sensor Simulation for Multi-Robot Scenarios

NVIDIA just pushed a major update to Isaac Sim, focusing on realistic sensor data for simulating robot swarms. This directly addresses the sim-to-real gap for collaborative tasks, letting you test coordination logic before you buy the second robot.

Events

ROSCon 2025 Schedule Revealed: Focus on Production-Grade Robotics

The ROSCon 2025 agenda is live, and it's all about taking robots from lab demos to production lines. Sessions focus on safety certification, long-term maintenance, and real-time performance, which are the exact hurdles that kill startups.