This paper illustrates the development and evaluation of a humanoid robot called Robovie. This robot is designed to interact with people, communicate and contribute in the society in the role of a "partner". This interaction is achieved by using a large amount of behaviours which grow with the help of a visualization tool which measures the amount of complexity in the system. Several experiments have been established to test the level o.f communication and interaction between the humanoid robot and the human object. The approach suggested had very good results and opens a new path in interaction between humans and robots.
Problem Statement
When creating a new system, it is important to measure its performance so as to see how it is progressing. Partner robots have to be analyzed in regard to how they psychologically affect humans, which causes a problem as such concepts cannot be measured as easily as the mechanisms of a robot. Questionnaires interrupt the communication between the robot and the human, so the researchers used an evaluation approach where human behaviours (distance, attitude) were used to determine their feelings towards the robot. This was done by comparing them to an established assessment. Since it is not generally known how to efficiently create a robot with human like features, the researchers suggested adopting a constructive approach as that way they could monitor and modify the robots progress until it becomes human like.
The researchers found some communication problems that they had to resolve: interruption and deviation. For example when humans have a conversation and the doorbell rings they would stop talking to respond and open the door, that's the kind of behaviour they wanted to adapt to the robot. To do that they created reactive transitions and reactive modules built in episode rules. With this way when the reactive transition is a part of the current position and the precondition rule is satisfied then the robot executes the current action and then moves to the next one
Contributions
The researchers developed a humanoid robot called Robovie that have sensors to interact with humans. For achieving better results the researchers implemented a "constructive approach" to understand the complexity of the robot which acts like a human and interact with humans every day. They established experiments with the robot in order to evaluate the behaviour and analyse the interaction with the human object.
Approach
This paper presents an exploratory approach for the development and evaluation of interactive humanoid robots. They have developed an interactive humanoid robot that has a human-like body as the testbed of this embodied communication. Furthermore, many interactive behaviors have been implemented. This encourages people to treat the robot as a human child. Our evaluation approach is to measure the body movement interaction between a humanoid robot and humans and compare the results with traditional subjective evaluation. In other words, our constructive approach-to continue implementation until the robot behaves more human-like-aims to establish the ideal top-down model of interactive humanoid robot through exploratory bottom-up construction. As well, it provides a testbed for analyzing human-robot interaction, which produces plenty of knowledge for building interactive robots
Methodology
They constructed a robot called Robovie, which had human like form and designed specifically to interact and communicate with humans. It had several sensors for vision, touch and hearing. It was 120cm tall with 40cm diameter and weighted about 40 Kg. It also had two arms, eyes and a mobile platform (2 driving wheels and on free wheel for moving). The 16 skin sensors were the important parts of the system, 10 more to the mobile platform for vision, microphones to hear the human object and 24 more sensors for avoiding collision with the obstacles. The eyes of the robot had "pan-tilt" devices used for controlling what the robot could see. They created sensitive skin sensors for the robot to enable interactions with the human object. In addition the robot had battery which could last 4 hours and after that the robot could recharge by finding close recharge points. Robovie had a Pentium III PC with Linux as an operating system for handling data, generating Robovie's software modules and producing complex behaviours. The behaviours adapted to the robot were first experimented and then implemented, so they would understand better the complex behaviours of a human object.
Different behaviours were executed as rules to provide better communication. A tool was used for making the implementation stage easier, called "Episode Editor" which demonstrated the behaviours and how the combine together visually. The main plan for this stage was to create modules for each situation the robot could be, classify the execution order of the modules with the rules for keeping transition in order and insert episode rules for more automatic transitions. For making the robot to interact better they designed it in an action reaction way, so when the human object acts towards the robot, the robot at the same time reacts and vice versa. They applied a method to the robot where it checks if the precondition is satisfied and then move on to executing a certain task. For example when the robot has a conversation with the human object about the weather for example the robot can check the internet automatically and then talk about it with the human but when the robot cannot connect the internet that means that the precondition is not satisfied so it cannot follow the conversation.
The part of the robot which helps on recognising behaviours identifies the human's behaviour and reacts accordingly. After that the robot records which are the results of the current action and move on to execute the next one. The researchers also used a reactive module to force the robot to look at the part of it body that is touched by the human, so when the robot interacted with the human at some point the human object touch the robot so the robot look at the point that has been touched. In the experiments 26 students were participated as the human objects 19 of them were men and 7 were women and their average age 19,9 years. At the first stage they were shown instructions about how to interact with the robot and they were given the opportunity to observe the robot in a room 7.7m.x10m. The robot as part of its nature started an autonomous conversation with the subjects and the interaction began. The subjects after completing this stage they completed a questionnaire to evaluate the robot's behaviour on the experiments.
Outcomes
The researchers managed to adapt 100 behaviours to the humanoid robot, 70 of them were behaviours showing interactions like kissing, hugging, hand shaking, greeting, singing songs, showing towards objects, having a conversation. Another 20 were "idling" behaviours like the robot scratching its head and moving its arms. The last 10 were "moving around" behaviours like keeping a watch and travelling around an area. The robot is "proactive" on starting interaction with the human object and waits for what the response action will be. The design of the robot makes it easy for the human objects to interact with and imitate through the experiments. According to how the human acts the robot chooses the best way to respond illustrating the interaction between them. The experiments highlighted that the robot had a very high performance when interacting with humans. They were acting with the humanoid robot like it was a human child and kept eye contact with it for more than half of the experiment's period. In addition some of the participants had a physical interaction with the robot. The physical design of the robot made it look like a human, and it's interactive behaviours also gave it human like qualities. The robot, with speech recognition was able to talk with the human subjects but at the level of a child. The humans that interacted with it using conversation were given instructions by the robot to do specific tasks where the robot was responding back correctly.
Related Work
Scasselati [3] created a humanoid robot to test and evaluate joint attention between human and robot. The basis on this development was to create complex system with a variety of behaviours in order to interact with humans. The robot was called Kismet which has the ability to show behaviours performed by humans like anger, awareness, sadness, fright and excitement. The robot had microcontrollers, a motor system to move the robots facial features so it can show the expressions, a system for showing emotions and behaviours. [3] After experimenting with Kismet the results were very promising. The robot showed significant clues of autonomous behaviour with successfully gazing and changing expressions according to what the human object interacting with it was doing
Matsusaka et al [4] produced a robot that can look at the person that communicates with. The developed robot was named ROBITA and the main role it had was to be an information servant and a partner to a human object. They experimented with the robot in a conversation between humans and tested what were the behaviours it showed. The results showed that the robot was able to identify which person had the most attention from all the participants and gathered the attention to that person since it was the more likely person to talk next. The robot also showed that when it was asked a question by a human object it was gazing at that person when it was responding. Nakadai et al [5] developed a robot that can track a person when talking. The robot had several sensors so it can interact with a human object and being able to track it and recognise it, and motors to help it with the movements during the interaction. They used a real-time approach to make the robot identifying the object and perform complex tasks when sound was detecting in the environment. They used a tool to observe what is going on in the environment in 3D and give to the robot the ability to have ID face recognition and automatically identify the human object that talks and also track various objects.
Recently other robots have been developed for interaction with human objects like Honda's ASIMO and Sony's AIBO which open a new path in robotics with their impressive results. Sakagami et al [1] created ASIMO, a humanoid robot which was autonomous and could interact in an environment designed for humans. The first samples of ASIMO are already in the industry and some of them are used as receptionists. Humanoid robots like ASIMO could be very helpful in everyday life for every human and can be used to perform many tasks in their life. They installed a navigation system to the robot to move around the environment easily, a \pc for speech recognition to help it through the interaction, a radio controller to be able to talk with the human objects, a DSP board to identify sounds and respond, microphones, sensors, camera to be able to look at the human and Ethernet. [1] The experiments illustrated that ASIMO can perform the follow: detect a human object, identify the face of the human, send him greetings, guide the human object, introduce the human, detect obstacles and avoid them. The human can also call ASIMO and when it confirms that is a known person it responds to that call and follows. In addition it can handshake, and when showed a direction it can go there and stay. Sakagami et al [1] used an architecture that had a main goal which had successfully implemented, to demonstrate that ASIMO can perform tasks helpful for humans in their everyday life.
Fujita [2] has developed a pet style robot which was designed to act as a real pet. They adapted to this pet robot many complicated behaviors and gave it complex behaviors which characterize a dog. This robot can be used as a companion for older people and as a game for children since it is designed to play football so it is also used in RoboCup Legged League. Experiments were held to demonstrate the robot's complexity and the various behaviors it can execute. Fujita [2] by developing this pet robot had made a very big step to a new era of autonomous robots that can interact with humans.
Conclusions
The researchers manage to illustrate successfully the construction of a humanoid robot called Robovie which had various sensors on its body to enable interaction with humans. They made experiments with the robot using university students who interacted with it like it was a human being. They were involved in many activities with the robot, observed its behaviour like gazing, or staring at them, playing with it, exercise illustrating the amount of body movements it was capable of doing and showing at the same time the high level performance and results of the robot through the interaction. After further analysis of these movements positive relationships between them and subjective evaluations were discovered. Nowadays humanoid robots are used in education and in so many other fields. One of the approaches used in the experiments for analysing the body movements is also significant in many tasks in embodied communication like developing the behaviours in robots and analysing human communication
