Another humanoid robot that mastered balance, is the ASIMO robot from Honda. Are the methods of keeping balance different from ATLAS? Let’s find out!

Figure 1: The ASIMO robot Source: ASIMO

Besides the fact that ASIMO is able to walk as we do, it also understands preprogrammed gestures and spoken commands, it recognizes voices and faces and interfaces with IC Communication cards. These IC cards use infrared signals to receive and transmit information. In this way, ASIMO can detect your presence even if you aren’t within the line of sight of its cameras. And it wouldn’t be a humanoid robot without its arms! ASIMO can do things like turn on light switches, open doors, carry objects, and push carts. ASIMO is 1.3 meters high, thus a bit smaller than ATLAS. The Engineers from Honda did a lot of research, for instance with looking at the legs of insects and mammals. Take for example the fact that we shift our weight using our bodies and especially our arms to make sure that we stay in balance. These findings were very important in getting ASIMO’s walking mechanism right. ASIMO even has soft parts on its feet that play a similar role to the one our toes play when we walk. What role do our toes play when we walk? They help your feet to bear the weight of your body. When running, your toes effectively increase the overall length of your foot, allowing you to run faster! The soft material also absorbs the impact on the joints. This is comparable to the soft tissues in our bodies, like for example our tendons. ASIMO is also using a gyroscope sensor to regulate the balance. It is mounted on its body and performs several tasks.

• The gyroscope senses the position of ASIMO’s body and the speed at which it is moving, relaying adjustments for balance to the central computer. This is similar to the way our inner ears maintain balance and orientation.
• As we now know, interaction with the environment plays a major role in the balance of bipedal robots. Through floor surface sensors in the feet and six ultrasonic sensors in its midsection, the robot can detect objects around itself. This is the same as ATLAS does with a LIDAR, but with a different kind of sensors. Then it compares the gathered information with maps of the area stored in ASIMO’s memory.
• In the video about the balancing organ, we came across proprioception. ASIMO has both joint-angle sensors and a six-axis force sensor to get this functionality. This way it accomplishes the job our muscles and skin do in sensing muscle power, pressure, and joint angles. With this information, ASIMO has a better understanding as to where its parts are located in the environment and can thus correct itself better when losing its balance.

 

In the field of balance, many little functionalities are big milestones. At first sight, you are maybe not very impressed by ASIMO just walking like a human. But think about the turning capabilities. Rather than stop and shuffle into a new direction, ASIMO leans and smoothly turns like a human. This is accomplished by finding a way to work with the inertial forces created when walking. For example; the earth’s gravity creates a force, as does the speed at which you walk. Together those forces are called total inertial force. When your foot connects with the ground, also a force is generated. This is the ground reaction force. All these forces need to balance out and posture has to work to make it happen. This is called the Zero Moment Point (ZMP). It specifies the point with respect to which dynamic reaction force at the contact of the foot with the ground does not produce any moment in the horizontal direction. Thus the point where the total of horizontal inertia and gravity forces equals zero.

The Honda Engineers worked on three areas of control, to control ASIMO’s posture:

• The soles of the feet absorb the unevenness of the floor while still maintaining a firm stance. This is called floor reaction control.
• The target ZMP control also plays an important role. When ASIMO can’t stand firmly and its body begins to fall forward, it maintains its position by moving its upper body in the direction opposite the impending fall. Simultaneously, a higher speed of walking is generated to counterbalance the fall.
• When the target ZMP control is activated, the length of the step is adjusted to regain the right relationship between the position and speed of the body and the length of the step. This is also known as foot-planting location control.

 

ASIMO can turn without stopping, which most robots cannot do! When we walk around corners, we shift our center of gravity into the turn. Predictive movement control, also called Honda’s Intelligent Real-Time Flexible Walking Technology or I-Walk, accomplishes the same thing by ASIMO. This real-time technology predicts how much ASIMO should shift its center of gravity to the inside of the turn and how long that shift should be maintained. With every step this bipedal robot takes, it has to determine its inertia and then predict how its weight needs to be shifted for the next step in order to walk and turn smoothly. The following factors are adjusted to maintain the right position:

• Length of its steps
• Its body position
• Its speed
• The direction in which it is stepping

 

The engineers of Honda started with ASIMO being able to walk, but that wasn’t enough for them. They wanted ASIMO to run. There is a big difference between running and walking regarding the contact with the ground. When walking, there is always one foot that is in contact with the ground. When running you jump and have a floating fase where neither of your feet touches the ground. Through coordinated arm and leg movements, you stay in balance when moving your feet in the air. While researching this, they encountered an entirely new set of challenges. ASIMO’s torso was given a degree of freedom to aid in bending and twisting so that the robot could adjust its posture while airborne. Otherwise, ASIMO would lose control while airborne.

Earlier it was mentioned that stereo cameras play a big role in the interaction with the environment. But how about shadows, odd angles, and movements that these counteract? Does ASIMO still work then?

The vision system of ASIMO consists of two basic video cameras for the eyes, located in its head. It uses stereoscopic vision, which is also used in the human brain. It translates the visual input of the cameras into a three-dimensional structure and creates a perception of depth. Together with a proprietary vision algorithm, it makes sure that ASIMO can see, recognize and avoid running into objects even if their orientation and lighting are not the same as those in its memory database. These cameras can detect multiple objects, determine distance, perceive motion, recognize programmed faces and even interpret hand motions. For example, ASIMO stops when you hold your hand up to it in a “stop” position. The facial recognition feature allows ASIMO to greet familiar people.

Several other sensors also help maneuver through environments and interact with objects and people. Surface sensors detect objects and changes in the floor. Ultrasonic sensors help by detecting surrounding objects. Discrepancies between the internal map of the area preprogrammed in its memory and the actual environment are resolved by these sensors.

Just like ATLAS, ASIMO isn’t an autonomous robot. It can’t make decisions on its own. It either has to be programmed to do a specific job in a specific area that has markers that it understands or has to be manually controlled by a human. Making ASIMO move by remote control may not seem that progressive. However, ASIMO does have a lot of other capabilities. It can for example self-adjust its steps. When it is walking forward and encounters stairs it automatically adjusts its steps to accommodate the terrain.

There are quite some similarities between ATLAS en ASIMO. Both use stereo cameras and gyroscopes to maintain balance. But what is the difference? ASIMO is scaled under the social robots, ATLAS however is an industrial robot.  The qualities and capabilities of ATLAS may rival social robotics in ‘impressiveness’, they are so different in their functionality that to consider them together would certainly be an uneven playing field. ATLAS is unique in the high mobility that it obtains. Unlike ASIMO, it has the ability to balance while performing tasks such as carrying items even when it is pushed. One of the major differences between ATLAS and more traditional robots is the use of hydraulic actuators instead of traditional servos. Hydraulics gives ATLAS much more power. Nonetheless, A major disadvantage is the loud and disruptive movements that come along with this.

Researchers have been looking into the balance mechanism for a very long time now. Hard work pays off: there are several bipedal robots nowadays. By using gyroscopes and accelerometers that can sense acceleration in angular and linear directions, they are aware of (unexpected) movements and able to counterbalance! Corrections in ASIMO are generated through target Zero Motion Control.

Now,  you are familiar with examples of balance mechanisms of bipedal robots. Maybe it’s possible to translate this into a working mechanism for exoskeletons? It definitely is something to think about…

 

Sources:

1. Wikipedia contributors. (2020, July 17). ASIMO. Wikipedia. https://en.wikipedia.org/wiki/ASIMO
2. Strickland, L. A. O. J. (2020, June 30). How ASIMO Works. HowStuffWorks.
   https://science.howstuffworks.com/asimo.htm
3. Honda Global | ASIMO. (2020). Honda Robotics. https://global.honda/innovation/robotics/ASIMO.html