Humanoid Robot (Shanghai) Co., LTD uses Xsens mocap and AI to train humanoid robots to mimic human movement and perform complex tasks.
Key takeaways:
The word robot was first introduced by Czech writer Karel Čapek in the 1920’s to describe creatures that looked and moved like humans but used only to perform hard labor. In recent history, the human mind has been captivated by robots, both in science and science fiction. Our pursuit of creating machines that mimic or surpass human capabilities to fulfill tasks, both mundane and complex, has brought us a long way.
The humanoid robotics industry is one of the fastest growing sectors in 2024, contributing a whopping $1.32 billion in market share as of today and projected to reach $38 billion in 2035. Major players such as Tesla, Ubtech, Nvidia, Boston Dynamics, and others have been giving us sneak peaks and flaunting their intelligent robots, quickly bridging what was once considered science fiction and today's reality.
In the near future, robots are expected to perform a broad range of tasks, from folding laundry to operating complex machinery. While household helpers and robot first responders are still in development, the gap between fantasy and reality is narrowing. These robots are becoming more intuitive, adaptive, and human-like.
Modern humanoid robots not only resemble humans more closely but also mimic human motion with remarkable precision. A key to this rapid advancement is their ability to learn from humans in real-time through a process often referred to as human shadowing. Motion capture technology plays a critical role in this process. This is how it works: a person performs tasks multiple times, while every detail of his movement is carefully captured using motion capture technology. In the process of teleoperation, a robot performs the same operation next to the person, literally learning how to move through recorded motion data, and, using AI and deep learning, incurring all the possible options for future movement. By analyzing data from human motion, these robots can be trained to improve their dexterity and coordination, making them more prepared for real-world applications.
Humanoid Robot (Shanghai) Co., LTD has been a prominent player in this field, gaining attention at the recent World Robot Conference. The company has developed several generations of highly bionic humanoid robots designed for tasks in the manufacturing industry, including product sorting and assembly. Its focus on precision, adaptability, and artificial intelligence positions Humanoid Robot (Shanghai) Co., LTD as a key player in meeting future demands across various sectors.
In the pursuit of building humanoid robots capable of replicating human-like actions, the company has chosen Xsens motion capture technology to train and refine its robots’ movements. This collaboration has enabled the company to make significant progress.
We sat down with the team at Humanoid Robot (Shanghai) Co., LTD to talk about its use of motion capture to train humanoid robots, the role of AI and what the future holds.
The importance of accurate motion data
Data precision is crucial in training humanoid robots, as it directly affects their ability to replicate human movements accurately. Even minor errors in motion data can lead to mishandling or damage, especially tasks requiring fine motor skills or in robot-human interactions. Any inaccuracy in the initial data can compound errors in learning, leading to inefficient or imprecise movements.
With this in mind, and after testing several alternatives, Humanoid Robot (Shanghai) Co., LTD selected Xsens inertial motion capture solution for the task. As the team explained:
"We tested a lot of different motion capture solutions, both optical and inertial. The optical solution had problems of space limitation, so we wanted to use inertial motion capture. With inertial systems there is a problem of drift – when the motion data doesn’t align perfectly with the real movement. Xsens turned out to be the only inertial solution that had no data drift issues."
The robots are currently being taught various tasks, such as handling small objects and walking with great precision. One demonstration of the robots' abilities is its fine work with object identification and transfer: "We can control the robot to achieve the fine work of sorting out red beans from soybeans."
AI and motion capture: A perfect partnership
The second critical aspect of the learning process for humanoid robots is Artificial Intelligence.
As noted by Humanoid Robot (Shanghai) Co., LTD, motion capture and AI are used in tandem to give robots the human capabilities to move, interact with their environment, make autonomous decisions, and perceive, recognize, and respond to external stimuli intelligently.
AI uses data from motion capture to create models that help robots learn how to interact with their surroundings. A key component of this process is building a large data pool for training—covering a wide range of movement patterns, responses, and problem-solving techniques. Xsens allows for unlimited scalability by enabling multiple workers wearing Xsens suits to simultaneously perform tasks. These tasks are then captured and replicated by humanoid robots, while creating a rich data foundation that allows robots to adapt and respond to diverse situations.
By scaling data collection with Xsens, robots can mimic human actions with remarkable accuracy while dynamically adjusting to different environments and tasks. Additionally, AI optimizes this motion data by learning from errors, enabling continuous improvement without human intervention.
As Humanoid Robot (Shanghai) Co., LTD explains, AI further streamlines its processes by automatically annotating collected data, significantly improving the efficiency of its training programs. The combination of AI and Xsens’ high-precision motion capture has allowed the company to push the boundaries of what’s possible.
Looking ahead: the robotized future
The company sees humanoid robots as a transformative force across industries. "Humanoid robots have a very large potential, and there will be a huge demand in the future in industry, business, and eventually in the consumer side. For example, many countries are currently facing the problem of aging populations and insufficient labor. Future humanoid robots can solve these problems well, and it can also create more new jobs and form a new industrial revolution." The company predicts that robots will not only fill workforce gaps but also create new jobs and revolutionize entire sectors.
Achieving this vision of the future starts with teaching robots how to be more human-like, and Xsens motion capture that has typical use in areas such as animation and healthcare, is now playing a vital role in teaching robots how to move like us, humans. With its high precision, scalability, real-time data processing, and freedom to capture movement in any environment, Xsens meets all the critical requirements for a tool to train humanoid robots.
Humanoid Robot (Shanghai) Co., LTD’s collaboration with Xsens highlights the potential of advanced motion capture in shaping the way humanoid robots move in the future. Combined with AI, Xsens motion capture is helping the company make massive strides in realizing a long held human ambition – creating robots in our own image.
Learn more about humanoid robot motion training with Xsens.