Skip to content

Home

Spatial World Model
Modeling interactions between people, objects, and robots in space
A new technology that can "imagine" future situations in space

Challenges in the Physical AI Field

SpatialWorldModel_01
Research and development of Physical AI, which enables AI to understand the physical world, is accelerating.
However, current Physical AI is mainly used in well-organized factories and warehouses. In offices and commercial facilities where people move freely, the movements of people and objects are unpredictable, making autonomous robot behavior difficult.

Solution with Spatial World Model Technology

SpatialWorldModel_02
We developed a spatial World Model that integrates information from cameras placed throughout the space and cameras mounted on robots. This model creates a comprehensive 3D scene graph representing the interactions between people, robots, and objects within the space. Based on this, we model the interactions between people, robots, and objects to predict future actions.
This enables autonomous robots to avoid collisions in advance and perform collaborative operations.

Value Brought by Spatial World Model Technology

  1. Advanced Autonomous Robots
    • Can avoid unintended collisions between autonomous robots, which was difficult from a single robot's perspective.
  2. Human-Robot Collaboration
    • Realizes smooth collaborative operations by estimating the intent behind human actions.
  3. Robot-to-Robot Collaboration
    • Achieves advanced inter-robot collaboration by coordinating robots within the space.
     This technology can be applied not only in factories and warehouses but also in various fields such as security and commercial facilities where there is a high need for labor saving and efficiency.

Technology Overview

Target Industries and Users

Industries and users requiring labor reduction in person-to-person services such as security and customer service

Challenges in Target Industries and Operations

  • While AI agents are streamlining operations in digital spaces, efficiency improvements through human labor in physical spaces have reached their limits, raising expectations for Physical AI. Although technology exists for individual robots to operate autonomously, multiple robots interfere with each other and fail to work properly.

Technical Challenges

  • Technology to execute autonomous operations of multiple robots in parallel without interference

Value Brought by Spatial World Model (Details)

The Spatial World Model enables labor reduction in physical spaces through human-robot collaboration by imagining collaborative operations between humans and robots.

Fujitsu's Technical Advantages

Research and development of World Model technology that gives AI imagination is actively underway. While other companies build World Models only from the perspective of individual robots, Fujitsu constructs a World Model of the entire space, including not only individual robots but also sensor data deployed in the space. This allows for a greater amount of imaginable information and enables coordinated control of not only individual robots but also multiple robots.

Usage Scenarios

  • End Users
    • Introduction of robots in situations requiring labor reduction
  • Application Developers
    • Add-in to security systems and other applications for autonomous coordinated robot control

Examples and Use Cases

  • Multi-robot control environment that can handle both visitor guidance and security in office, public facilities and commercial facilities