Through the official Bonsystems Global channel, a Wheeled Quadruped Robot equipped with the BCSA V4 actuator has been unveiled. This robot is not merely a demonstration model, but a platform designed with real operational environments in mind, built upon Bonsystems’ long-accumulated expertise in actuation technology and structural design.
The robot represents the result of carefully identifying the essential elements required to operate high-output actuators reliably and integrating them seamlessly into a unified system. Its structure, drive system, and balance are organically connected, reflecting a design philosophy that prioritizes practical, on-site application throughout the entire platform.
During the design process, the experience gained through extensive testing and validation served as a key foundation. Considerations such as load distribution, structural stability, and the optimization of drive efficiency are embedded across every aspect of the Wheeled Quadruped Robot. In particular, the overall architecture is harmoniously configured to ensure stable and effective utilization of the BCSA V4 actuator’s performance.
The robot is offered in two models to accommodate different operational requirements. Model 107 is designed with a focus on fast responsiveness and mobility, providing a well-balanced platform for environments that demand dynamic motion and efficient movement. Model 127, on the other hand, emphasizes high output and stable load handling, featuring a reinforced structure and drive system suited for industrial settings and heavy-duty tasks. While distinct in character, both models share a common technological philosophy centered on real-world deployment.
Built around the BCSA V4 actuator, this Wheeled Quadruped Robot is intended for stable operation across repetitive tasks and diverse environmental conditions. The direction Bonsystems Global has consistently pursued—creating robots that can be trusted in the field takes on a more tangible form through this latest release.
For those interested in the underlying technology and real-world operation of the robot, more detailed information and demonstration footage are available on the official Bonsystems Global YouTube channel.
Looking ahead, Bonsystems Global will continue to introduce robotic platforms that deliver practical value to industrial environments, guided by its accumulated technology and hands-on experience.
Discover how reinforcement learning trains humanoid robots to interact with the real world. Learn the difference between supervised, unsuper
🤖 Reinforcement Learning: How Do Humanoid Robots Learn on Their Own?
Humanoid robots are no longer just pre-programmed machines. They are rapidly evolving into Physical AI systems that learn through experience, much like humans, by interacting with the real world.
This article explores Reinforcement Learning, the core technology behind humanoid robot training, and explains how robots acquire human-like movement through trial and error.
Key Highlights
✔️ Three Machine Learning Paradigms: Supervised Learning, Unsupervised Learning, and Reinforcement Learning
✔️ What is Reinforcement Learning?
→ A learning method where robots explore their environment, receive rewards or penalties, and gradually discover optimal actions
✔️ How Humanoid Robots Learn Safely
→ Training complex motions in virtual physics environments using the Sim2Real (Simulation to Reality) approach
✔️ What Makes Physical AI Possible
→ The integration of perception through sensors and vision, decision making with AI models, and execution via actuators
✔️ Why Hardware Matters
→ Durable, high performance actuators are essential to withstand the repetitive and high impact nature of reinforcement learning
✔️ Imitation Learning vs. Reinforcement Learning
→ Learning by copying demonstrations versus discovering solutions through autonomous exploration
Reinforcement Learning is more than just an algorithm.
It is a fundamental requirement for humanoid robots to operate reliably in the real world.
🔗 Read the Full Article
Learn how reinforcement learning, Sim2Real, and hardware reliability are shaping the future of humanoid robotics.
👉 [Read the full article on Reinforcement Learning and Humanoid Robots]
I would like to introduce a video that explains, in a very clear and intuitive way, how a cycloidal gear with two disks operates. Cycloidal reducers are widely used in robotics, automation equipment, and various precision control systems, and this video offers an accessible look into their essential structure and working principles.
The video begins by outlining the basic components of a cycloidal gear. The system consists of a sturdy housing and an input shaft. Notably, the input shaft does not rotate the disk directly. Instead, an eccentric cam mounted on the shaft creates a slightly offset rotational motion. Because of this eccentricity, the disk does not rotate smoothly around its center; rather, it produces a characteristic wobbling motion. This oscillatory movement is then transformed back into a smooth rotational output through the internal mechanism, enabling the gear to deliver high torque and precise speed reduction.
A particularly helpful part of the explanation is why two cycloidal disks are used in most designs. The video makes it clear that this dual-disk configuration helps reduce vibration and minimize torque ripple. Torque ripple refers to uneven transmission of force similar to how a bicycle pedal feels harder to push when it is vertical and easier when it is horizontal. By arranging two disks symmetrically, the vibrations generated by each disk cancel one another out. As a result, the system produces a much more uniform output, which is crucial for applications that require smooth and accurate motion.
The video also briefly explains how the gear ratio is determined. The number of lobes on the cycloidal disk defines the reduction ratio, and the final output motion is generated through output pins that fit into holes in the disks. In the example shown, the input shaft must rotate 49 times for the output to rotate once, and the output turns in the opposite direction of the input. counterclockwise when the input rotates clockwise.
This video is an excellent resource for anyone wishing to understand the fundamentals of cycloidal gear mechanics, even without prior familiarity. If you are curious about how mechanical motion is transformed into high torque and stable rotational output, I highly recommend watching the video and following along with the explanation.
Recommended Reading
For those who would like a deeper understanding, I also recommend the article titled “Cycloidal Gearbox Design: Principles, Structure, and Working Mechanism.” It provides a more structured explanation of the design principles and structural characteristics of cycloidal reducers, and when viewed together with the video, it will help clarify the overall operating mechanism even further.
Bonsystems Global has released a new video that explores why many humanoid robot companies are choosing to develop cycloidal reducer technology internally, and how Bon Systems’ own BCSA V4 cycloidal reducer represents a core breakthrough in this direction.
The video opens by highlighting the growing attention toward BCSA V4 in the robotics market throughout 2025. It also revisits the Robot World 2025 exhibition, where Bonsystems unveiled its original humanoid robot and a four-wheel, four-leg mobile platform equipped with this new actuator architecture. Rather than treating actuators as simple components, the video frames the technology as a fundamental structural philosophy behind next-generation robotics.
One of the central themes is the company’s current focus on absolute hardware completeness. The video explains that robust hardware attracts the interest of software companies seeking reliable platforms for advanced control solutions, leading to active collaboration discussions. This perspective reveals how BCSA technology operates not just as a mechanical part but as a foundation for integrated robot development.
The breakdown of Unit 2, the humanoid robot used for real locomotion and manipulation tests, is another key highlight. Different BCSA V4 lineups are used across the robot according to functional demands.
For the upper body, where space constraints are severe, the compact yet highly durable RI series (070, 087, 096) is used. These actuators enable natural shoulder and arm motion while maintaining a slim, humanlike profile.
In contrast, the lower body must support the robot’s full weight and generate significantly higher torque. Here, the RO series (107, 127) is applied to deliver powerful output through an extremely thin profile, maximizing torque density and structural efficiency.
Around the midpoint, the video features a demonstration of back drivability, where a person manually moves the robot’s arm to show how softly and smoothly the joints respond. This behavior not only enhances human-robot interaction but also contributes to safety by absorbing external impacts.
The video also explains how Bonsystems integrates its own driver and encoder into BCSA V4 while still allowing users to replace them at any time. This flexibility supports a wide range of system environments, and it ties directly to one of BCSA’s strongest advantages: customizable gear ratios tailored to application-specific requirements.
Toward the end, the company shares its engineering goals for next year. The upper body is planned to adopt teleoperation and imitation learning, while the lower body will progress toward full bipedal walking. These steps outline a clear roadmap for BCSA-powered humanoid evolution.
The video concludes with a preview of the next episode, which will take a closer look at the four-wheel, four-leg platform expected to enter real industrial environments before the humanoid itself. As the two systems develop in parallel, they form an ecosystem of locomotion and manipulation technologies built on Bonsystems’ cycloidal reducer expertise.
