Terms

Vision-Language-Action Model

Follow the latest coverage, related explainers and connected technology stories.

Latest coverage

CN
Destro AI Raises $8 Million to Build a Layer Coordinating Robots and Workers in Warehouses

Destro AI Raises $8 Million to Build a Layer Coordinating Robots and Workers in Warehouses

Destro AI has emerged from stealth with an $8 million seed round after developing Vision and Mothership systems to coordinate robots, workers, and trucks in logistics-handling operations. The company is expanding an initial pilot with Yusen Logistics to 26 robots, alongside another pilot involving 17 robots in Southern California.

CN
Japan’s GENIAC Program Expands Funding for Robotics AI Models

Japan’s GENIAC Program Expands Funding for Robotics AI Models

Japan’s GENIAC program, supported by the Ministry of Economy, Trade and Industry and NEDO, announced results and funding for projects to develop foundation models for robots and build a data ecosystem for embodied AI. The initiatives include 13 organizations in the robotics-model field and nine organizations in the data-ecosystem project.

CN
Will the Large Behavior Model LBM Become the Next Core Architecture for Autonomous Driving?

Will the Large Behavior Model LBM Become the Next Core Architecture for Autonomous Driving?

MONOist examines the shift in interest in artificial intelligence for robots and autonomous vehicles from VLA models to the Large Behavior Model (LBM), which focuses on generating behavior directly in the physical world. Toyota Research Institute and XPeng are highlighted as examples of organizations adopting this trend, while its broad practical viability remains under evaluation.

CN
Ricoh Develops Physical AI Technology to Teach Robots from Images of Human Actions

Ricoh Develops Physical AI Technology to Teach Robots from Images of Human Actions

Ricoh announced the development of technology based on a Latent Action Model to teach robots actions such as grasping and moving objects from image and video data, without relying entirely on a specific robot architecture. The company tested the technology in a simulation environment built on physical conditions, using AWS services to reduce the cost of data collection and model training.