Search button in the site


New Mobility Enabling
Highly Efficient Locomotion

Sony has developed new mobility robotics technologies capable of traversing rough terrains
unsuitable for wheel-based transportation.
The robots achieve stable, energy-efficient locomotion through a combination of Sony’s
proprietary technologies and are expected to find applications in a variety of industries, from
entertainment to construction and last mile logistics.

Yasunori Kawanami / Yasuhisa Kamikawa /
Masaya Kinoshita

A New Mobility Technology That
Expands the Possibilities of

We are working on developing “Tachyon”, a robotics technology for stable and energy-efficient locomotion. If robots were capable of efficiently transporting objects across uneven terrains, it would open up a range of possibilities in a variety of industries from entertainment to industrial applications, bringing about a major change in society. Based on this vision, we began research and development on new mobility mechanisms in 2015.

By combining Sony’s technological assets with the unique perspectives of our engineers, we have developed a variety of proprietary technologies in the areas of mechanics, control, and electricity.

Show video
Show video
Show video

Announced in December 2021, Model 3 features a six-legged wheel configuration

Click here for the description of the video content.

Demonstration of a six-legged wheeled robot in operation. The video shows that a mobile robot moves in a straight line, turns, and slaloms by the wheels and moves up a step by using both the legs and wheels.

A Series of Evolving Development

The development of robotics requires not only the improvement of individual component technologies but also the overall optimization of these technologies as a whole. We have been researching and developing new mobile robots and their component technologies, setting development goals, and upgrading our models in line with industry trends and changing market needs. Three models with different features have been developed and unveiled so far and each has built upon the core elemental technologies and design concepts of the last.

The robots’ name, Tachyon, which has been passed down through the three models, is derived from the name of a hypothetical particle that travels faster than the speed of light. It expresses the engineers’ desire to create a new mobile robot that can reach its destination speedily, regardless of whether the terrain is flat or uneven.

Model 1:
A fast-moving model built for
higher speed

The initial model is a quadruped with non-linear gears and a light-weight, high-power motor. It is capable of walking at a speed of 5 km/h over an uneven terrain (with variations of 5 cm) and preventing a fall under external force up to 40 Ns as well as has the ability to recognize obstacles and plan routes. In Model 1, we established basic component technologies such as legged locomotion, posture stabilization control, recognition integration technology, and local path planning.

Tachyon 1

Development period:
October 2015 to March 2018
Compact four-legged configuration / Non-linear gears / Lightweight high-output motors
Elemental Technologies:
Basic legged mobility technology / Posture stabilization control technology / Recognition integration technology / Local path planning technology / EDLC / Regeneration technology
A fast-moving model built for higher speed

Model 2:
A high payload model built
for transporting things

A quadruped like Model 1. In addition to the technologies of its predecessor, it features springs built into the legs to increase payload and energy efficiency. Vibration control technology is also introduced to suppress vibrations caused by the elasticity of the springs. Mode 2 is highly impact-resistant and capable of transporting loads of 20 kg—about half of its own body weight—in environments with stairs.

Tachyon 2

Development period:
December 2017 to February 2020
Four-legged configuration / SEA (elastic element) installed / Caster wheels can be mounted on the leg tips
Elemental Technologies:​
Basic technology inherited from Tachyon1 / Vibration suppression control technology / High load and shock resistance / Integration of legs and wheels / Low latency, multi-axis synchronous control system​
A high payload bearing model built for transporting things

Model 3:
A demonstration model for further
stability and energy efficiency

The latest model features a six-legged wheel configuration that combines wheels and legs, providing superior stability to the four-legged models. It incorporates the design principles of Model 2, including the use of springs to conserve energy when traveling by wheels*. The goal is to achieve highly efficient transportation while maintaining the payload.

* Some of the technologies are not yet available in Model 3 (as of January 2022)

Tachyon 3

Development period:
October 2020 to present
Six-legged wheel configuration / Structure for continuous stability / Efficient locomotion
Elemental Technologies:
Basic technologies inherited from Tachyon 1 and 2 / six-legged wheel drive technology / Continuous stability control technology / Multi-camera support Path planning technology
A demonstration model for further stability and energy efficiency

The following is an introduction to the key points of development in the areas of mechanics, electricity, and control, as well as the proprietary technologies.

Mechanical Technology for
High Payload and Energy

A compact, high performance leg

There are three major mechanical challenges when designing robots with legs. In the course of developing each model, we devised mechanical mechanisms to solve each of them.

The first challenge is reducing the size of the drive source. The force and speed generated by the legs depends on the angle of the joint. When a leg is bent, it is difficult to generate force, and when a leg is extended, it is difficult to generate speed. This issue was solved through the structure of the legs. In Model 1, the comma-shaped nonlinear gears and Three-link legs equalized the leg movements. Model 2 adds a linear-motion, miniaturized, unequal-length four-bar linkage.

The second challenge is to achieve a high payload while conserving energy. Even simply standing up consumes energy. High reduction ratio gears can be used to reduce energy consumption, but these gears tend to break due to landing impacts. Therefore, Model 2 uses high reduction ratio gears and Series Parallel Elastic Actuator (SPEA) with a shock-absorbing structure. Model 2 and 3 achieves to reduce energy consumption through introducing springs parallel to assist the power output of the motor.

The third challenge is the effect of the leg-joint popping out and the risk of interference between legs when climbing stairs. In the case of a link-type legged robot, there is a risk of collision with items in the surrounding environment when climbing up and down stairs. For this reason, some quadruped robots climb down stairs in reverse. There is also risk of collision between front and rear legs, which necessitates a longer, thinner body. To eliminate the risk of interference and reduce unnecessary movement, Model 3 uses a leg with a linear actuator. The two-stage telescopic structure allows for lightweight legs and compact expansion and contraction. We are also developing a small customized motor with even higher output than Model 2.

A compact, high performance leg design

Changes in leg configuration

The newly developed Series
Parallel Elastic Actuator

SPEA (Series Parallel Elastic Actuator) was a key technology to achieve the high payload for Model 2. In general, SPEA has been used in biped robots at the research level before but we developed a unique structure to make it smaller and lighter and implemented it in a quadruped robot for the first time during the Model 2 development period.

There are two conventional methods for driving the legs with a motor. One is the springless proprioceptive actuator used in Model 1, but this requires a larger motor and more current. The other is using a Series Elastic Actuator (SEA) in parallel with a spring. The SEA senses the deformation of the spring and sends feedback to the encoder to move the legs. The issue here is that a larger mechanism is required to combine the springs. In contrast to these methods, the SPEA has a structure that combines two types of springs, one in parallel and one in series, in the actuator. The series spring absorbs the shock applied to the legs at high reduction ratios, preventing damage. In addition, the force applied to the actuator can be sensed through the deformation of the spring, which can be used for force control. The parallel springs assist the output of the motor.

From left to right, it is composed of the parallel spring, ball screw, linear slider, motor, series spring, and output rod. The entire actuator moves from side to side with the nut of the ball screw as the center of rotation. The figure on the left shows the movement when the motor is activated. The entire linear actuator moves to the right. The movement of the output part of the actuator on the right end is decelerated through the four-bar linkage mechanism to drive the lower limbs. Since the nut is fixed in place, less load is placed on the linear slider, thus allowing for a reduction in size and weight. Furthermore, by placing the motor, which should be left of the center of rotation, on the right, the length of the ballscrew is reduced and parts such as support bearings are no longer needed.

In addition, the two small series springs are designed to detect the slightest force displacement using a linear encoder capable of detecting displacement over a distance of 0.2 μm.

The newly developed Series Parallel Elastic Actuator

Electrical Technology for Stable
and Powerful Operation

Predicting and managing the total
power output of the motors

Legged robots are usually equipped with several high-power motors. If the total output exceeds the power supply capacity, a power flicker will occur, causing the robot to stop or become unstable. The options are thus to either increase the power supply, which means making it larger, or to limit the total motor output, which means powerful operations will no longer be possible.

To solve this dilemma, we built a system in which the CPU controlling the motion predicts the total output required for the target motion based on the instructed torque and rotation speed of each actuator and corrects the motion if it exceeds the power supply capacity. Rather than uniformly limiting the power output, the output is controlled for each individual actuator while managing the overall motion of the robot, thereby reducing the risk of power flickers while maintaining powerful performance.

EDLC (Electric Double Layer
Capacitor) for instantaneous peak

Furthermore, legged robots require instantaneous high power output to drive the legs, such as when walking or jumping. So, the high power battery must either be larger or have smaller capacitance. That is why we decided on an EDLC (Electric Double Layer Capacitor). This is an electric storage device with a lower capacity but a much higher output than a battery.

Incorporating an EDLC supports peak power while minimizing weight gain, thus also contributing to raising the payload of each model. Model 2 was able to achieve a carrying capacity-to-weight ratio of 0.5, which is one of the highest in the world.

Electrical Technology for Stable and Powerful Operation

By using an EDLC for instantaneous power output and batteries for average power output, we have already assessed a peak power of 4000 W in Model 1. For Model 3, we are aiming for double the peak power of the battery.

Control Technology to Maintain
Stable Operations in Changing

Whole-body cooperative control
and multi-contact
control as the cornerstones of

The superiority of our method lies in the fact that it simultaneously achieves multiple motion objectives. This makes it possible to operate with the back in a horizontal position even when external forces are applied. We believe this will be a major advance for practical applications such as transporting cargo. The key is the whole-body cooperative control and the multi-contact stabilization control.

In order to control the whole-body cooperative motion of a robot, it is necessary to find the driving force for each joint to realize multiple motion objectives. For computation purposes, we use our in-house developed Generalized Inverse Dynamics (GID) system. By imposing constraints on the amount of force that each joint can produce and an upper limit on external force generation and calculating the desired driving force for the entire body within the range of those constraints, it allows the robot to move without straining itself. Sony’s GID boasts a high calculation speed and is versatile enough to be used for a variety of robots.

The other technology, multi-contact stabilization control, allows the robot to make contact with the surrounding environment while simultaneously using the external force to achieve the desired motion and stabilization. Even when the ground is less even than expected, or when the robot’s body is pushed sideways due to contact with a person, the robot can respond nimbly at a near-human speed to avoid falling over and continue its original motion. The stabilization control’s outputs are the leg’s contact point with the floor, the current CoP (Center of Pressure), and the gait cycle. For example, if the robot is pushed to the side while walking forward, and its feet cannot reach the ground, it stabilizes itself by autonomously changing its walking cycle. In terms of stabilization control, the strengths of Sony’s system include high versatility regardless of the number of legs, support for running and jumping in addition to walking, fast computation time, and the ability to autonomously change the walking cycle.

Model 1 has a low-inertia mechanical configuration that allows for agile movements through a combination of whole-body cooperative control using GID and multi-contact stabilization control, meaning that it does not require torque sensors to measure joint force. In Model 2, we were once again able to achieve torque-sensor-less force control (SEA is used only for the knee joints) despite the large gear ratio and rigid mechanical configuration. We were also able to improve the robustness of the multi-contact stabilization control to allow the robot to carry unknown loads.

Whole-body cooperative control and multi-contact stabilization control as the cornerstones of stability

Toward the Pinnacle of
Mobility Technology

To further improve these technologies, we are also accelerating our efforts to verify the roles and technologies required of mobile robots. Model 3 is being developed in collaboration with Shimizu Corporation, and we have started demonstration tests using the robot at construction sites. In addition to evaluating locomotive performance, we are also hoping to upgrade the robot based on needs in the field. For example, the robot may need to avoid areas such as puddles or openings. Or it may need to move through spaces with glass or mesh walls, which are difficult to recognize, or in spaces exposed to strong sunlight.

In addition to that, one of our major visions of the development is “coexistence of humans and robots,” so we are also working with Sony’s Creative Center on a design that, while multi-legged, will not make people feel uncomfortable. In this way, we are continuing our development while imagining how this technology could be put to practical use in the future.

Show video
Show video
Show video

Click here for the description of the video content.

Image video showing the demonstration experiment by a six-legged wheeled robot. It introduces the operation of robot moving around obstacles such as stepladders and moves up and down the step at the construction site.​
Show Image

(Left)(Above) Video of a demonstration experiment at a construction site
(Right)(Below) Possible future design under consideration with the Creative Center

We are aiming for the pinnacle of mobility technology. If we can realize mobility technology that enables a robot to travel all kinds of terrains with excellent energy efficiency, the application of robots may be expanded not only to daily life scenarios like last mile logistics, but also to extreme scenarios where people cannot go, such as mountainous areas and outer space.

Through the development of practical mobile robotics technologies, we will strive to provide unprecedented new value to users in a variety of fields.


Show Image

Yasunori Kawanami

Tokyo Laboratory 24

Through many years of research and development, Sony has developed a wide range of technological assets, from hardware such as actuators, sensors, devices, and materials, to software such as control, environmental recognition, and path planning. We are engaged in the development of cutting-edge mobile robots, and it is not inconceivable that we will invent something that the world has never seen before. Our hope is that we will develop technologies that will be useful in various industries and people’s lives around the world.

Show Image

Yasuhisa Kamikawa

Tokyo Laboratory 24

Sony has an attractive research environment where talented specialists and engineers can work together in friendly competition to develop new products. Developing robots entails a lot of responsibility and continuous learning, but I’ve been working in development with a big picture perspective from a young age and making use of my own ideas. I also find it rewarding to collaborate with external partners and vendors. I hope you will join us in creating the world’s greatest robotics technologies.

Show Image

Masaya Kinoshita

Tokyo Laboratory 24

We research and develop robots based on a comprehensive strategy, from planning to design, assembly, and commercialization. Since we are a small team of specialists, we are able to experience every aspect of robotics development. In addition, the benefit of working at Sony is that if we cannot solve a problem, we can call on the help of experts in many other fields. I would like to continue working with colleagues who are not bound by convention and who are passionate about changing the world of robotics.

to the top