Robots can tell directions at stations, introduce menus at restaurants. These images are used to be the stories in the future. But, they have become reality.
Hitachi has developed the robotics IT platform and the humanoid robot "EMIEW3" that can support humans' daily activities by its independent action from thinking for itself. This technology has already been applying to our society practically.
As to achieve this technology development, many young researchers have dedicated their passions and hard works on how to make robots to co-live with humans.
Researcher: YAMAMOTO Akihiro, ASANO Yu, TOKUHASHI Kazumasa, and MATSUDA Tomohiro
(Publication: November 16, 2016)
Robot body development
YAMAMOTO Akihiro
YAMAMOTOEMIEW3 is the robot that is developed for providing services such as greeting to customers or giving guidance in public spaces like stations and airports as well as private owned businesses. Most of robots are used in industries. The purpose of EMIEW3 is to assist our daily lives and to live with humans. As to achieve the friendliness of EMIEW3, its body is designed with the round shape, and its voice is programmed to be child like.
We also focused on the size. The latest model is the third generation from the beginning of this development started in 2005. The body of the previous models was relatively large. In order to make communication easier with human, we reduced the size to 90 cm and 15 kg to bring its eye contact to the same height as a person sitting on a chair. With this compact size, EMIEW3 will not cause any harm when it bumps into a person.
Fig.1 Standing up after a fall
YAMAMOTOEMIEW3's biggest feature is to use those four wheels on the foot to move around freely. Its job is to offer information to guide people. EMIEW3 is designed to be capable of conducting movements smoothly not only in straight lines but also in direction changes. Furthermore, EMIEW3 is capable of picking itself up from any directions without assistant after falling over.
Movement control was really challenging in this project. We looked at ourselves at the mirror to identify how the body is placing its weight when we bend over and is balancing itself when we stand up. These observations were reflected to the robot control.
YAMAMOTOExactly. Even to move on the flat places requires the complex control, unfortunately the actual application will not limited to the places without any obstacles.
As to respond to the tactile paving surface and uneven roads, we spent a lot of time on developing the sensing technology for avoid obstacles and the technology to move on uneven road surface. It also requires to prepare multiple movement patterns for different situations such as how to take the next action when move among the crowd. Currently, EMIEW3 is capable of overcoming floor level differences in 15 mm.
System development
MATSUDA Tomohiro
MATSUDAEMIEW3 is composed of the robot body and the computer system with cloud. The smaller the robot is, the less processing capability of the build-in computer gets. To solve this problem, we designed external application platform on cloud for voice recognition, location detection, language processing and so on, therefore, only essential function requiring real-time response such as collision avoidance runs in robots. We call this platform "robotics IT platform." By this means, we designed EMIEW3 with smaller and lighter body which can behave responding to voice recognition result and realize accurate movements.
Fig.2 Overall structure of EMIEW3 and
Robotics IT platform
MATSUDAFor example, EMIEW3 can extend translation services for 100 languages and update software easily. Also, we can control EMIEW3 remotely in case of disaster or emergency and monitor multiple EMIEW3s located in different locations at once.
EMIEW3s can share destination information and offer a seamless service as to guide customers who are asking directions in the airport. EMIEW3s can collaborate each other to provide sophisticated service.
MATSUDAI have taken the role of developing the operation and monitoring system of EMIEW3 for one year. Before joining this project, I was involved in another work for different technology field. In this project, I have to collaborate with team members with unfamiliar technology such as robotics and voice/video processing. Additionally, I have less experience for work in team, therefore, I spent much time to get used to work in this EMIEW3 project.
If system has to deal with too much data, it might face technical trouble such as system down. We chose essential data to be sent and solve this problem, however, various kinds of problem often occurs. In such case, all the related engineers make discussion and try to find solution by trial and error manner. Either technical solution or communication sometimes requires much time, however, we have been asking questions each other for finding solution and moving forward step by step.
AI processing development
ASANO Yu
ASANOEMIEW3 is a robot that is capable of co-living with humans. How to communicate with humans without any problems is the most important feature in this development. EMIEW3 can start a conversation to the person who seems to be in trouble. This action requires three types of processing that are the "image processing" to detect people who has stopped, the "voice processing" to convert the recognized voices into the text form, and the "language processing" to response through proper phrases.
For example, the functions requires EMIEW3 to be able to support customers who are looking for products at the electronics stores are (1) finding the person who is in trouble, (2) understanding the question, (3) explaining where is the product, how much is the product, and what are the functions of that particular product. All three must be integrated adequately to provide satisfied customer service.
Fig.3 Robust voice processing in ambient noise
ASANOLet me think. It should be how to make EMIEW3 talk like humans. Humans conversation do not limit to questions and answers but include greetings such as "how are you doing?" There will be various patterns of response to follow this type of conversations. It was not easy to collect all the patterns of response to anticipated questions. Moreover, even the response is proper to casual situation, sometimes it could be unsuitable phrases for customer service. The brush up of whether conversation or response is appropriate will be needed for advancing EMIEW3.
During actual conversation, there would be announcements, other people's conversations, and crowd's noises. We also developed the technology that can extract the conversation of target person from the crowd.
ASANOExactly. The integration of three processing, image, language, and image, was the key features of this development. Even each individual process has 90% processing capability independently, it is very often that the capability drops to 70% when three are integrated. It was quite difficult to increase its overall accuracy as well as its individual accuracy. However, this is also where I felt enthusiasm for.
Business strategy
TOKUHASHI Kazumasa
TOKUHASHIPrevious EMIEW2 was focusing greatly on publicizing its technology development. Now, the technology of EMIEW3 is to aim at developing a robot that can be apply to daily lives to provide service in stores with a business purpose.
As for the usage, it will be mainly for the customer or information service. There It is also possible to offer other applications with each technology independently. We are exploring the possible needs, and the method of promoting those needs. In the future, I would like to cultivate this technology to become a profitable business as a pillar of the company.
TOKUHASHIBy looking at EMIEW3, customers probably have difficulty to actually understand how Hitachi has utilized its service on robots. If customers are not able to feel the benefits, the business will not be successful. We need to first listen to customers for their requests carefully. It is also important to show demos to them for actual realization. From these, we will explore the possible applications and services.
The next stage will be the detailed validation experiment. Sometimes this is conducted at the customers' location. In our Akasaka office in Tokyo, we have a space that can perform prototyping with customers collaboratively through developing the direction of concept and service, and discussing specifications. Also, we have a facility called "Robotics Co-creation Room" in Ibaraki to conduct validation experiments.
After all these steps, the developed technology will be deployed to the market. Currently, we have several projects are in progress.
TOKUHASHIThe staffs who are engaging in technology development need to not only conduct their researches over the desk and in the lab but also have the communication ability to listen from customers for what they want and how that can be achieved by EMIEW3, and to express their own thoughts. I joined Hitachi as a researcher. However, I am also interested in business strategy and business development. This is how I became involved in the business strategy of EMIEW3.
I feel there is the similar excitement between research and business in the point of how to find the solution for the challenges.
YAMAMOTONow this type of technology development has been progressed all over the world. Without cautions, other companies will develop new technology and we will become less competitive in the market. Although I have felt quite pressure during my research due to its significance in Hitachi's businesses, I also have experienced great satisfaction & accomplishment when I achieved something. Moreover, I am very honored to see the robot that I designed is actually working in society.
MATSUDAThere was no manual for this development. Everything was based on our own ideas and speculations with multiple attempts. We were not able to anticipate the outcome. While repeating trial and error, I have enjoyed greatly in this project to develop a new technology. This experience will be accumulated for my career. Through this project, I have learned so much that I will value for the rest of my life.
ASANOTo explore the possibilities of robots to perform tasks that human can not complete is interesting for me. For example, when the command is to "memorize 1 billion words by tomorrow," robot could be capable of carrying out the task easily but humans are not able to do it. I would like to develop a robot that will support humans and make contributions to society.
TOKUHASHIThis project has allowed me to challenge many different tasks beyond the research and to consider of planning, proposing, and calculating profit for business model. All these were the valuable experiences to me. Since this development, we have been working on the concept of co-living with humans. I would like to promote this to become a profitable business as a pillar of Hitachi while raising its recognition in society.
EMIEW:Excellent Mobility and Interactive Existance as Workmate