Forcesteed Robotics develops "Forcesteed-LEIVOR," a general-purpose humanoid controller designed to accelerate the real-world deployment of humanoids.
Forcesteed Robotics has developed "Forcesteed-LEIVOR," a general-purpose humanoid controller designed to accelerate the deployment of humanoids in business environments. This innovative framework merges traditional robot teaching methods with AI robotics technologies like imitation learning and inference, enabling robot SIers to leverage their existing expertise for humanoid implementation. The controller aims to bridge the gap between conventional development and AI robotics, facilitating easier adoption and practical use of humanoid robots in various industries.
📋 Article Processing Timeline
- 📰 Published: April 3, 2026 at 19:00
- 🔍 Collected: April 3, 2026 at 10:30
- 🤖 AI Analyzed: April 18, 2026 at 01:02 (350h 31m after Collected)
Forcesteed Robotics Co., Ltd. (Headquarters: Koto-ku, Tokyo; CEO: Hiroyuki Osawa; hereinafter "Forcesteed Robotics") is pleased to announce the development of "Forcesteed-LEIVOR," a general-purpose humanoid controller.
Forcesteed-LEIVOR is a framework that integrates AI robotics technologies, such as imitation learning and inference processing, with traditional motion design through "teaching," a method widely utilized in the industrial robot sector.
By enabling the integrated management of robot **L**earning, **E**xecution, **I**ntelligent Behavior, and **O**peration (**LEIVOR**), the framework promotes the introduction and operation of humanoid robots while leveraging the development expertise of existing robot system integrators (SIers).
Future plans include expanding support for LeRobot VLA (Vision-Language-Action) models and integrating the system with image recognition technology and previously announced robot picking packages. Currently, the controller supports robots from ugo, RealMan, and Unitree, with plans to progressively increase the number of compatible models.
### Background and Objectives
While the humanoid robot field is expected to see significant growth, actual deployment in business environments remains limited. One primary reason is that the development and introduction of humanoid robots have not necessarily aligned with the methodologies used by the robot SIers who have traditionally handled industrial robot implementation.
In the world of industrial robotics, program design via "teaching"—where a human manually guides the robot's movements—is standard. Conversely, recent AI robotics often employs approaches that generate robot behavior through imitation learning and inference. This divergence in development philosophy has made it difficult for existing robot SIers to directly apply their technical skills and operational know-how to the humanoid field.
To bridge this gap, Forcesteed Robotics developed Forcesteed-LEIVOR as software capable of handling both traditional teaching-based development and AI robotics within a single implementation flow. The goal is to move humanoid utilization beyond a few advanced players and R&D purposes, making it easier for field teams to conduct Proof of Concept (PoC) trials with a view toward practical application.
### Key Features of Forcesteed-LEIVOR
* **An Extension of Industrial Controller Philosophy:** Forcesteed-LEIVOR is positioned as an extension of the integrated controller philosophy from the industrial robot domain into the humanoid field.
* **A General-Purpose Platform:** Rather than being software dedicated to a specific humanoid model, it is designed as a "general-purpose humanoid controller" (a hybrid of traditional programming and imitation learning) applicable to various humanoid and semi-humanoid robots. It aims to serve as a common development platform rather than having fragmented environments for each robot.
* **Leveraging Existing SIer Expertise:** Users can advance motion design and business implementation for humanoids by utilizing the teaching and programming concepts cultivated over many years in the industrial and collaborative robot sectors. This lowers the barrier for existing robot SIers to enter the humanoid field.
* **Integration of Imitation Learning and Inference:** The system can incorporate imitation learning using human work data and inference processing based on specific situations. This enables robot operations in environments where teaching alone is insufficient or where site conditions change dynamically.
* **Flexible Operation:** By combining robot control utilizing VLA models with traditional rule-based control, flexible and adaptable operations become possible.
### Future Outlook
To implement humanoids in society not merely as individual machines but as "field-adaptive general-purpose robot systems," Forcesteed Robotics will continue to expand support for VLA models such as LeRobot. Furthermore, we will promote the construction of robot systems capable of enduring real-world operations by integrating image recognition technology and existing software packages.
Forcesteed-LEIVOR is a framework that integrates AI robotics technologies, such as imitation learning and inference processing, with traditional motion design through "teaching," a method widely utilized in the industrial robot sector.
By enabling the integrated management of robot **L**earning, **E**xecution, **I**ntelligent Behavior, and **O**peration (**LEIVOR**), the framework promotes the introduction and operation of humanoid robots while leveraging the development expertise of existing robot system integrators (SIers).
Future plans include expanding support for LeRobot VLA (Vision-Language-Action) models and integrating the system with image recognition technology and previously announced robot picking packages. Currently, the controller supports robots from ugo, RealMan, and Unitree, with plans to progressively increase the number of compatible models.
### Background and Objectives
While the humanoid robot field is expected to see significant growth, actual deployment in business environments remains limited. One primary reason is that the development and introduction of humanoid robots have not necessarily aligned with the methodologies used by the robot SIers who have traditionally handled industrial robot implementation.
In the world of industrial robotics, program design via "teaching"—where a human manually guides the robot's movements—is standard. Conversely, recent AI robotics often employs approaches that generate robot behavior through imitation learning and inference. This divergence in development philosophy has made it difficult for existing robot SIers to directly apply their technical skills and operational know-how to the humanoid field.
To bridge this gap, Forcesteed Robotics developed Forcesteed-LEIVOR as software capable of handling both traditional teaching-based development and AI robotics within a single implementation flow. The goal is to move humanoid utilization beyond a few advanced players and R&D purposes, making it easier for field teams to conduct Proof of Concept (PoC) trials with a view toward practical application.
### Key Features of Forcesteed-LEIVOR
* **An Extension of Industrial Controller Philosophy:** Forcesteed-LEIVOR is positioned as an extension of the integrated controller philosophy from the industrial robot domain into the humanoid field.
* **A General-Purpose Platform:** Rather than being software dedicated to a specific humanoid model, it is designed as a "general-purpose humanoid controller" (a hybrid of traditional programming and imitation learning) applicable to various humanoid and semi-humanoid robots. It aims to serve as a common development platform rather than having fragmented environments for each robot.
* **Leveraging Existing SIer Expertise:** Users can advance motion design and business implementation for humanoids by utilizing the teaching and programming concepts cultivated over many years in the industrial and collaborative robot sectors. This lowers the barrier for existing robot SIers to enter the humanoid field.
* **Integration of Imitation Learning and Inference:** The system can incorporate imitation learning using human work data and inference processing based on specific situations. This enables robot operations in environments where teaching alone is insufficient or where site conditions change dynamically.
* **Flexible Operation:** By combining robot control utilizing VLA models with traditional rule-based control, flexible and adaptable operations become possible.
### Future Outlook
To implement humanoids in society not merely as individual machines but as "field-adaptive general-purpose robot systems," Forcesteed Robotics will continue to expand support for VLA models such as LeRobot. Furthermore, we will promote the construction of robot systems capable of enduring real-world operations by integrating image recognition technology and existing software packages.