- Kenji Hashimoto, Assistant Professor (May, 2015)
Making Robots Run
My research focuses on bipedal robots that can run like humans. At the same time, I am studying disaster response robots leveraging the research results.
Research and development of bipedal robots at Waseda University began in the 1960s. We are trying to understand humans “constitutively” through the development of humanoid robots. A constitutive understanding means an approach to understand a subject not by observation or analysis but rather by actually creating and moving it. For example, through trial-and-error research on humanoids that can walk in a human-like manner, we have found that the movement of the pelvis is important for humans to be able to walk. In addition, enabling humanoids to walk in a human-like manner has contributed to other fields of study such as the quantitative assessment of the usage of welfare devices.
Although we have already succeeded in making robots walk in a human-like manner, if we are to try to make them run, then we have to overcome a huge hurdle. In order to run, the robots must use their legs to jump, and this motion requires significant power. Robots gain force mainly from motors, but if you were to use only a motor to create the force necessary to jump, the motor will be too big to be contained in the size or weight of a human. Additionally, as humans do not gain force from their muscles alone, this will deviate from the purpose of making it run “like a human.”
In fact, the elasticity of muscles and tendons is also used when humans jump. When we analyze running motion, we see that energy during landing is stored in elastic parts and is released to produce a large force. I am therefore currently working on incorporating the same mechanism into a legs-only robot. We are replacing muscles and tendons with motors and springs. We took steps starting with a model that employs coil springs as the leg structure, and then advanced to a structure with human-like joints and leaf springs. As of now, we use a side support to prevent the robot from falling, and are considering improving the ankles in order to remove the support by the end of AY 2015.
Figure 1: Bipedal robot with leaf springs in the knees and ankles (Courtesy of Assistant Professor Kenji Hashimoto)
After this model, we are going to add a trunk, arms, and sensors to make it more human-like, but we believe that the state where both feet are floating during running would pose a big challenge. How to handle a situation where the robot is about to fall when its feet are in the air will be difficult, and we anticipate that movement of the arms and the rest of the upper body and coordination with the lower body will be important. In order to run fast, we focus on how to move the arms. There have already been robots that can run, but if we look at them closely, their arms do not move.
This research is conducted jointly with Professor Yasuo Kawakami in the School of Sport Sciences at Waseda University. This is because we need to understand the movement of the human body in order to control a robot to make the same movements as a human, using the same mechanisms in the same manner. Another reason is because it is supposed that we can use robots that run like humans to develop a reasonable coaching method for athletic runners.
Developing a Quantitative Coaching Method
According to sports science researchers, a quantitative analysis of how to run faster in a short-distance sprint has not yet been performed. Although there have long been analyses of the running form of fast runners, we suppose that the use of robots will make more efficient analysis possible. We can start testing from extreme movements that humans cannot make, and we will not have to worry about injuries due to overloading. We are going to take an approach where if we find something that might make the robot run faster, we will adjust the movement so that it can be performed by humans. We believe that if we can shorten the short-distance sprint time with this approach, we will learn a secret way for humans to use their entire bodies more effectively. It would also be possible to apply this to other areas of athletics. In long-distance running, for example, we have two major running techniques—forefoot and heel-strike—and there are various opinions as to which is more efficient. Using robots can allow us to conduct quantitative validation of this problem.
According to our research prospects, it will be ready for use in training athletes who will compete at the Tokyo Olympic Games. Robots may also be used for purposes other than training validation, such as exhibition races with athletes and as torch runners.
Path to Disaster Response Robots Working in Extreme Environments
I am also involved in the development of disaster response robots based on past research achievements. After the Great East Japan Earthquake and the Fukushima nuclear disaster, there have been heightened expectations for robots that perform rescue and other operations in areas that humans cannot reach because of the danger of collapse or high radiation risks. The Cabinet Office of the Japanese government formed a research and development program called the “ImPACT program” in such an environment, and I am in charge of the development of legged robots in the program. In this capacity, I am working mainly on quadruped robots that crawl, and I feel that I can widely apply the know-how we have gained from our research on bipedal walking.
The advantage of legged robots is their mobility in structures, such as vertical ladders, spiral staircases, and stairs with uneven step heights. Other types of robots are not so good at this. For bipedal or quadrupedal walking, however, the problem is recovery after falling. We are considering a method of movement that minimizes the risk of falling over by making the robot crawl on its belly with four legs like a crocodile. Also, in case of failure, the robot must be able to move, by itself, to a location where it will not obstruct the activities that come next. For this reason, the legged robot we are developing has been designed to be capable of movement even if two of its four legs are broken.
Figure 2: A rendering of the disaster response legged robot under development (Courtesy of Assistant Professor Kenji Hashimoto)
Several disaster response robots have been developed in the past, but we have repeatedly seen that they were not usable in actual disasters due to lack of maintenance. Therefore, in this project, we are proceeding with development using a concept that aims for robots that can be used for maintenance of aging infrastructure or plants in non-emergency times. As the maintenance of aging tunnels and bridge columns is becoming a problem, there is likely a necessity for this kind of robot.
Lastly, I think it is critical that a gap in the understanding of robots is developing between experts and laymen. Many people would associate the word “robot” with a humanoid. However, my understanding of robot, as an expert in this field, includes all devices that are controlled by controllers in response to information received by sensors and moved by actuators. In my definition, remotely controlled heavy equipment is a kind of robot, and so are automatic doors. I feel that there is another similar gap in the capabilities expected of robots. In the disaster in 2011, robots were not able to accomplish results to the extent that people expected. I think that we must improve the capabilities of the robot we are currently working on to the level where it can be of use in actual sites.
Interview and Composition: Noel Kikuchi
In cooperation with: Waseda University Graduate School of Political Science J-School
