A New Era of Embodied Intelligence: Multi-Modal Bipedal Robot Enables Advanced Mobile Manipulation

Imagine this: a dual-wheeled bipedal robot steadily climbs a staircase while firmly grasping an object with its robotic arm. The transition is seamless, the movements precise. It’s a demonstration that not only breaks away from traditional flat-surface constraints but also redefines mobile manipulation by extending it into complex 3D environments.

Conventional mobile manipulation is typically limited to planar environments where robot arms operate within constrained workspaces. This new robot platform defies such limitations. With its hybrid locomotion capabilities—supporting point-foot, bipedal, and wheeled modes—it expands operational boundaries into stairs, narrow corridors, and irregular terrain, all while maintaining manipulation precision.

In a synchronized demonstration, two robots work in tandem: one holds a box steady, while the other accurately places an object inside. This level of coordination resembles a well-trained duo and showcases the platform’s potential for collaborative multi-agent tasks.

The robotic arm is capable of grasping from various angles, further hinting at future applications in dynamic environments such as household assistance. Its integration with a multi-modal mobile base unlocks new research directions for real-world deployment scenarios.

What’s particularly noteworthy is the platform’s minimalist design—a robotic arm mounted on a dual-mode leg system (wheeled and bipedal). This simplicity drastically reduces the complexity of simulation and algorithm design, offering an accessible gateway into mobile manipulation research.

Unlike high-DOF humanoid robots, which often present steep learning curves, this platform enables students and researchers to focus on core algorithmic challenges without being overwhelmed by complex dynamics. It effectively lowers the barrier to entry while preserving research depth.

Most existing robotic manipulation studies are restricted to predefined tabletops and static environments. In contrast, this new mobile manipulation kit is purpose-built to overcome those limitations. It allows researchers to design tasks that span multiple rooms and involve varying terrains—scenarios that more closely mirror real-world applications.

This enhanced terrain adaptability paired with fine manipulation capabilities makes the platform especially relevant for fields such as home service robotics and search-and-rescue, where unpredictability and terrain complexity are the norm.

Another key feature is its support for both decoupled and whole-body coordinated control modes. In decoupled mode, locomotion and manipulation are treated as independent modules. In coordinated mode, full-body strategies can be explored—for example, leveraging lower-body motion to stabilize or assist upper-body actions during manipulation.

This facilitates complex task execution, such as extending reach or maintaining balance when handling heavy objects. Researchers can explore how body dynamics influence task performance, making it an ideal testbed for full-body coordination studies.

Equipped with multiple peripheral interfaces, the robot can be quickly configured for a wide array of research applications. From 3D mapping with LiDAR and depth cameras to human-robot interaction via voice interface modules, the platform offers researchers the flexibility to explore diverse areas such as:

Perception and environment modeling

Navigation and dynamic obstacle avoidance

Sensor fusion and spatial awareness

Its extensibility ensures that researchers from various disciplines—AI, robotics, human-computer interaction, and more—can use a single platform tailored to their specific needs. This kind of flexibility empowers interdisciplinary innovation and fosters integrative research workflows.

From Mobile Base to Integrated Research Platform

Originally introduced as the world’s first hybrid bipedal robot featuring point-foot, bipedal, and wheeled configurations, this platform has now evolved into a comprehensive research tool. With the recent introduction of its robotic arm kit, it has transformed from a mobile base into a fully integrated mobile manipulation platform.

Each iteration and hardware refinement has been grounded in real-world lab testing—over 1,000 days of experimentation have led to a robust system with high stability and a user-ready out-of-the-box experience. The development philosophy has always focused on usability, reliability, and relevance to actual research challenges, rather than flashy gimmicks.

As embodied AI continues to rise as a key research domain, having an accessible, capable, and expandable research platform is more critical than ever. Analysts predict that the global embodied intelligence market will grow from $1.8 billion in 2023 to $13.8 billion by 2028, with a compound annual growth rate of over 50%.

This platform addresses a core need: giving researchers a powerful yet approachable tool that supports everything from foundational locomotion and manipulation to integrated perception-planning-action pipelines. Rather than spending time debugging hardware, researchers can focus on advancing algorithms and solving meaningful problems.

Ultimately, this new robot exemplifies what modern research platforms should aspire to: modular, extensible, reliable, and above all—aligned with the real needs of the research community.