In this blog post, we’ll explore the reasons for creating humanoid robots with forms similar to the human body, examine the control, path planning, joint, and hand manipulation technologies required to achieve this, and take a look at how far current robotics technology has advanced.
When people think of robots, they often picture the bipedal machines featured in the movie ‘The Terminator’. However, the robots that are actually having a significant impact on our lives are industrial robots performing precision tasks in factories. These robots have provided great convenience to our lives by reliably performing repetitive tasks that require precision. Nevertheless, we continue to strive to create humanoid robots with forms and functions similar to those of humans. So why are we so determined to create humanoid robots? It is because robots that mimic human physical structure and movement have an advantage when working in hazardous environments on behalf of humans or when utilizing spaces and tools designed for humans. Recently, research and industrial efforts to develop general-purpose robots based on human movement have been gaining momentum, and there is a growing trend toward utilizing humanoid robots in manufacturing and logistics environments specifically designed for humans. So, what kind of technology is needed to build a robot like the one in “The Terminator”? Let’s examine this from a mechanical engineering perspective.
The motivation behind building bipedal robots stems not only from human curiosity but also from concerns for human safety and quality of life. There has been a steady demand for mobile robots capable of performing tasks on behalf of humans in environments too dangerous for people to enter directly, such as nuclear power plants or locations with toxic gases. In fact, humanoid robots for disaster response have been developed to perform tasks such as operating valves or flipping switches in environments that are difficult for humans to access. Humanoid robots like NASA’s Valkyrie were also developed with the potential for use at disaster sites or in environments where human activity is difficult. This is because, in living and industrial environments designed around humans, robots with a human-like form can more easily utilize existing facilities and tools, such as stairs, doors, workbenches, and tools. Therefore, a humanoid robot is not merely a machine that mimics human appearance, but rather a robot capable of performing various tasks on behalf of humans in environments created for them. So, what technologies are required for such robots?
First is robot control technology. For a robot to walk safely without falling, it must plan the movements of each joint and control itself to precisely follow the generated trajectory. In particular, bipedal walking requires much more complex control than wheeled robots or fixed industrial robots, as it involves continuously shifting the center of gravity while supporting the body on two legs. Traditionally, PID control has been widely used for this purpose. PID control is a feedback control method that combines three actions—P (proportional), I (integral), and D (derivative). It measures the output of the controlled object, compares it to a desired reference value or setpoint to calculate the error, and adjusts the control input based on that error. This ensures that the robot’s joints follow the target position and velocity. However, in current humanoid robot control, PID control alone is not sufficient to handle all movements; instead, an approach is used that combines joint control with dynamic models, whole-body control, sensor fusion, optimization techniques, and learning-based control. In particular, for dynamic movements such as walking, it is necessary to simultaneously control multiple joints while monitoring the robot’s posture, center of gravity, and ground contact status in real time. Therefore, to ensure that a robot can stably perform the movements desired by a human, technology is needed that goes beyond simply moving individual joints to provide integrated control of the robot’s entire motion. This is why mobility, manipulation capabilities, sensor integration, computer vision, object recognition, control algorithms, and power and control systems are all cited as core technologies for humanoid robots.
Second is the technology for planning a robot’s path. The most ideal robot should be able to independently detect surrounding obstacles and actively plan a walking path suited to the situation. However, planning the gait of a bipedal robot is extremely difficult because it must account for the robot’s high degrees of freedom and complex configuration space. In complex environments containing various types of obstacles, finding feasible movements for a bipedal robot becomes even more difficult, and finding a safe yet efficient path in a high-dimensional configuration space requires significant computational resources and time. Furthermore, since the robot must avoid obstacles without falling, simply finding the shortest distance is not sufficient. Factors such as the robot’s balance, joint ranges of motion, potential foot placement, ground conditions, and the possibility of collisions with surrounding objects must all be considered together. To solve these problems, information from various sensors—such as cameras, depth sensors, LiDAR, and inertial measurement units (IMUs)—must be integrated to create a 3D representation of the surrounding environment, and the movement path and gait must be planned based on these results. NASA’s research on disaster response robots has also utilized technology that employs stereo cameras and LiDAR to map the surrounding environment and navigate complex terrain. Meanwhile, biomimicry also provides inspiration for robotic locomotion. For example, bats use ultrasound to detect obstacles in their surroundings and navigate; this approach applies the sensory and locomotion principles of such organisms to robots. However, today it is not only important to directly mimic the senses of specific organisms but also to combine various sensors—such as cameras, LiDAR, radar, and inertial sensors—to enable robots to perceive their surroundings.
Finally, technology is needed to replicate the human form, particularly the precise movement of joints and hands. The human hand produces extremely subtle and complex movements through the organic interaction of numerous joints, muscles, and tendons. Therefore, faithfully replicating the human hand and joints in a robot is by no means an easy task. While humans can control extremely subtle forces and movements, robots find it difficult to achieve the same level of precision due to limitations in actuators, materials, gearboxes, sensors, and control technologies. Furthermore, even if it is possible to move each joint individually, grasping or manipulating an object while moving multiple joints simultaneously involves complex interactions of force and torque between the joints. If unexpected twisting or impact occurs during this process, it can cause damage to the actuators, joints, or the robot’s structure itself. Therefore, when designing a robotic hand, rather than simply replicating the human hand exactly, one can also simplify the structure to make it suitable for performing the required tasks. For example, instead of making the number of robotic fingers exactly the same as that of a human, one can select the minimum number of fingers and joint structures necessary to grasp and manipulate objects stably. This approach helps ensure the necessary functionality while reducing the robot’s weight and complexity.
The technologies discussed above are just a few of the core technologies that make up humanoid robots. Stable bipedal locomotion requires control technology, dynamics, and sensor technology; navigating complex environments requires environmental perception and path planning technology; and performing tasks similar to those of humans requires sophisticated joint and hand manipulation technology. Only when technologies such as power supply, actuators, lightweight structures, artificial intelligence, visual perception, and human-robot interaction are appropriately integrated can a practical humanoid robot be created. Research robots like CIROS ver3.0, developed in the past by the Robotics Laboratory at Seoul National University, can be viewed as significant examples in this process of technological advancement. At the time, this robot was designed to move its arms and joints—excluding the legs—and to recognize and act on human commands; however, it had the limitation of being vulnerable to terrain obstacles because its legs consisted of wheels. Nevertheless, this type of research served as a process of exploring how to replicate human physical structure and the environments humans inhabit in robots. Currently, humanoid robot research is evolving beyond the simple mimicry of form seen in the past toward the ability to move in real-world environments, manipulate objects, and collaborate with humans. The International Federation of Robotics also explains that the potential for utilizing humanoid robots in manufacturing and logistics environments designed for humans is expanding, while pointing out that, for application in actual industrial settings, they must meet requirements such as reliability, efficiency, energy consumption, maintenance costs, safety, durability, and consistent performance. Furthermore, Seoul National University recently officially launched its Robotics Research Institute, presenting “robots for humans” and “human-centered physical AI” as its primary research directions. While humanoid robots continue to advance in this way, perfectly replicating human movement and capabilities remains a challenging problem. In particular, simultaneously achieving human-level walking stability, precise manual manipulation, situational awareness, energy efficiency, and safety within a single system is extremely complex. However, as robotic actuators, sensors, artificial intelligence, control technologies, and materials science continue to advance, the range of applications for humanoid robots—capable of assisting human activities or performing certain tasks on their behalf—is expected to expand even further in the future.
