Among the factors described above, more attention are given to action, human-robot interaction as well as architecture control of the robot.
A. “Human-oriented perception” by interpreting human activities and identifying human feedback;
B. “Natural human-robot interaction” which addresses normal social responses of dialogues or actions;
C.“Readable social cues” that could reflect its own internal statues back to the user through expressions, gestures or words; D. “Real time performance” to provide users with timely responses (Fong et al, 2003).
Two major functionality for the product are identified as motion and display, as both would assist design considerations for the level of interactivity.

Design thinking should be taken into consideration during various stages of product development, including product functionality, apparel, interactive experiences, its emotional resonances as well as social factors. Product morphology includes designing of its form, shape, colour, materials etc. Though the robot’s morphology is suggested to be necessarily inline with its functionalities (Fong et al, 2003), the two factors co-contribute to the user’s interactive experiences; upon which the emotional values as well as social factors are revealed.

A steady servo (MG995) attached to a rotatory base in combination of two more servo motors were chosen to construct the robot inner skeleton (Appendix B: Prototypes). Selection was mainly based on resources available for faster prototype construction. This allows Babble to have three degrees of freedom: rotation of the entire body; rotation of its head in two directions, allowing it to nod and wave. It has a screen in replacement of a physical face.

Base & Servo #0. This is the base servo, the motor for the entire body rotation. Origin of the ground coordinate rests on plane 0, located at the robot base center. Direction of axes are the same as the coordinate directions for servo #0. Rotation axis for servo #0 is Z- axis, as such, it does not need to overcome external torque imposed by gravity at any position. Thus, there is small load requirement for servo #0. The base where Servo #0 is attached to has a stable structure; it has a for-legged support that assembles three metallic plates at the same time. One of the plates mounts servo #0.

Body & Servo #1. Servo #1 accommodates the nodding motion of the robot head, which also encounters the problem of excessive loading. The rotational axis for servo #1 is in plane 1, the XY-plane of servo #0. Little deflection of the servo arm results in significant loading on servo #1. Gravitational force accounts for the largest portion of the external torque. This directly affect the servo performance, namely, the robot falls over upon sudden changes of motion or electrifying; the structure could not stand straight if the weight above servo #1 is unbalanced. Though the problem is resolvable through changing servo #1 to one with larger torque inertia, the replacement is hard to accomplish in this case, as the entire base section was pre- assembled such that the structure is only suitable for the size of this type of motor.

Head & Servo #2. Servo #2 extends interactivity to a larger scale. The robot head could now sway left to right. The rotational axis is orthogonal to the screen. With servo #2, the robot head now has full rotational capability installed. By default, the position of the robot head has 25 degrees of extension, as shown in the figure above. This is under the consideration that for most of the interaction, the user’s eye sight comes from a level above the robot head, an extension would allow the users to view the screens easily.

Controller Board & Battery The controller board available for usage was pcDuino v2 board. This is similar to a personal-computer, where programs and commands can be loaded to perform desired operations. The operating system running on the pcDuino was Ubuntu 12.1 by default. In order to support the software requirement by the learning framework, it was updated to Ubuntu 14.1. The pcDuino has a small dimension of 10.5cm by 5.3cm, easily to be incorporated to the existing robot structure. The controller board, as well as battery component are most likely to be incorporated to the base shell segment of the robot. The space between robot base to the base of servo #0 is currently designed to be filled with plastic translucent material, both for decoration and supporting purposes. Rechargeable battery could be incorporated within the base structure as an extra weight for stability, and the controller board could be secured on top of the battery.

Screen The screen was selected as a 7’’ Capacitive touch screen (RobotShop, 2016) to allow touch input from the users. (Appendix C: Babble Prototype) However, due to the limitations of no existing driver available for enabling touch function on pcDuino, the previously designed touch interactivity is achieved through reading user’s input from the keyboard and respond accordingly. The back of the screen is assembled to servo #2 through a single servo, due to the unavailability of servo connectors. However, the structure is not steady. Thus, a supporting “I” frame is proposed to be attached to the bottom of the screen. Subsequently, the screen and the frame is assembled to the head shell through small screws and a circular ring frame, and to be further enclosed with organic glasses mounted to the top of its surface (Figure 3.3a). Camera Module Another component not included in the drawing is the camera module. However, it has been incorporated during the testing stage. As seen from figure 3.3 above, there is a cavity above servo #2 behind the screen frame in its head compartment. This is where the camera as well as controller boards for the camera and the screen will be inserted. The plan for the next step is to fill in the cavity with a proper sized and good quality camera module.

The servos are controllable in two ways — through a motor control program or through sending in discrete line commands. While testing through the motor control program, dragging the pointer position could change the rotation of the servo motors respectively.
A set of actions were determined for the robot through this program. Actions including “casual looking”, “smile/happy”, “nodding”, “look left” and “look right” etc. were incorporated.

In order for the motors to rotate while connected to pcDuino, line command is needed for motion control. The format for a line command is:
#0P1500T1000 /r/n
The respective position of each servo at robot initial position:
Servo #0: P 1500 Servo #1: P 1130 Servo #2: P1580

If the desired action is a nodding action: //Initial position #1 at P1130
#1P1280T1000 /r/n
#1P1050T1000 /r/n
#1P1130T1000 /r/n

This could be easily interpreted. Range of action for for nodding action is to control servo #1 from P1050 to P1280. Starting from its initial position, servo #1 would go up 150 steps to P1280 in 1s, then move downward to P1050 in 1s, then return to its initial position in 1s.

To list an example for a more complicated smiling action:
//Starting from Initial position
#0P1600T700
#1P1080T700
#2P1700T700 /r/n

//Moved to designated position in 0.7s
#0P1600T700
#1P1080T700
#2P1700T700 /r/n

//Pause at designated postion for 0.7s
#0P1500T500
#1P1130T500
#2P1580T500 /r/n

//Return to initial position in 0.5s

Do notice that “/r/n” only occurred after sending in command for all three servos. The commands for multiple servos are interpreted by the controller as one whole command, which could be processed at the same time. The replication of code from the first to the second command block segment is because the action requires the robot to pause at the designated position for a while, in order to make the effect more realistic. There the pause, robot is returned to initial position.

Animate CC by Adobe is capable of creating 2D animation, and interactively programmed and controlled through JavaScript. The application was priorly know as Flash. The current version is updated with Web Graphics Library (WebGL) and HTML5 export compatibility. Though unable to create 3D models, it could add gradient to shapes instead to create depth to the animation. The biggest advantage of adopting Animate CC is its broad compatibility across browsers.

Expressions are designed and programmed in Animate CC. The expression animations are edited by adding key frames on the timeline. Animation sequences are transit and controlled by Javascript languages and pre-defined libraries. As touch function is disabled for the capacitive screen, keyboard is used as an input to trigger its animation responses. For example, if letter “a” is detected on keyboard,
//a=angry
if(e.keyCode===97)
{ this.gotoAndPlay("angry"); }

the system would go and play “Angry” sequence of the animation.

97 is the keyCode for letter ‘a’ on MacOS keyboard. This commands the timer to go to the keyframe assigned by name “angry”. In our animation sequence, keyframe 55 is assigned as “angry”.
Frame 0 is assigned as “idle” in our program. If letter “a” is pressed on the keyboard, the animation will jump to keyframe 55 and start playing “angry” sequence. The sequence ends at frame 106.

Frame 106 has a command of this.gotoAndPlay("idle");

such that once the timer reaches frame 106, it will read the command and return to play “idle” sequence starting from frame 0. “idle” is designated as blinking motion at static status. If no action is triggered, the system will play idle sequence in a loop.

Main body compartment of the robot is driven by servo #0 along with its head. Head of the robot is also controlled by servo #1 and #2. With reference to ground axis, the head of the robot could perform roll (sway), pitch (nod) and yaw (turn) motion (the whole body moves along servo #0 for yaw motion).

The expression is explained by screenshots of keyframes. Principle of disney’s animation rules were also implemented here for a more realistic effect (Stanchfield, 2009), most typically reflected in expression “Happy” and “Wired”.

As seen from the illustration above, the figures are elastic and vivid. “Wired” is designed to express a feeling of awkward or surprise, or when there are too many things to learn that it is unable to handle. As the eye-bulbs approaches each other very rapidly, they look like sticking together, being squeezed and bouncing back to the side again.

The “Happy” expression is also interesting. As seen from above, isolating the last two thumbnails may lead to an impression of arrogant or angry. However, while playing in a smoothed sequence, the expression looks lovely and projects a satisfying smile.

Thus, it is significant to view the expression sequence as a whole instead of isolating any key frame and look at them individually.

The choice of material closely affects the user’s responses on the product. As an example, shiny surface of plastic (ABS) material correspond to cheerful and modern outlook compared to those of a dull surface. However, researches have found that smooth plastic surfaces is correspondence to slippery perceptions, a feeling of sickness (slightly), coldness and easily contaminated with dirt or sweat. (Zuo, H, Jones, M and Hope, T, 2005). Thus, material selection rely on feedback from customers for more accurate choices.

The face of the robot will be covered with acrylic. The material has similar display as glasses, yet lighter in weight and higher impact resistant. Breakage of such material is unlikely to produce much fragments as when glass breaks, such that it is safer under such extreme circumstances.

The product is aimed for children who has less developed knowledge base. Target users have age range of 5 to 11. The product should be able to meet the users’ requirement. From the data of World Health Organization, girls from this age range have a height of 109 - 145cm, and boys from 110 - 143cm on average (WHO, 2016).
An estimation of 45 - 60 cm of arm length is estimated for age group of 5 - 11, correlates with the data from TUDelft.
These data matters with the design of the robot especially when it comes to human- robot interaction. Overall, the robot is 40.8 cm in height. At its natural position, the screen surface is 65 degrees to horizontal plane, and center of the screen is 29.8 cm above robot base.
At standing position, the child height ranges from 110cm to 145cm, desk height ranges from 55~65 cm for children of this age, and 65~75 cm for adults. The optimum angle for a single view position ranges from 10 degree above standard line of sight in the horizontal direction, as well as 45 degree below (MIT, 2011).

Structurally, Dew includes two major compartments – physical motion system and interactive I/O system. Motion system is where physical movement determines responses and interactions. Both its body and top-cap are capable of rotation. Dew is able to track and follow position of users through changes in its orientations. The mobility is achieved through two servomotors, one installed at the base that controls body-motion, the other at the body-cap intersection to control rotation of its tip. The interactive I/O system is where other functionalities are processed. Input system includes two stereo cameras as well as various sensors distributed throughout the robot body. Output system includes LED headlight, LED lining as battery status indicator, as well as LED interior for expressional responses; microphone for audio responses; motion also assists in expressing its responses. To ensure the output responses are emotional, Dew would be incorporated with a software engine that is able to analyse user’s expression and gestures, process the data from sensors and cameras simultaneously, and give a correspondent response through the musical, motional and expressional reactions.
Two main types of sensors are incorporated, including positional and bio- sensors. Position sensors calculates how the user interacts through changes in Dew’s relative position. More accurate data could be generated by combing the input captured through the camera. Bio-sensors include temperature, skin conductivity (SC) and blood volume pulses (BVP) sensors.
Researches have found that unimodal sensors is not accurate in detecting emotion- related variations because the result varies among individuals and the would fluctuate on a single person with one modality (Kim, 2007). This is the reason why multiple sensors should be adopted.
Sensors are chosen based on relative effectiveness and accessibility. As the most common body area the users would use to interact is their hands; data obtainable from hands is primarily considered. Moreover, SC is commented as the most effective measure of emotional-physiological relationship. Temperature, heart rate and pressure are also directly related especially with high arousal of users’ emotional states (Wagner, Kim, Andre , 2005 ́ ). Hot spots for user input is proposed as top tip and near-bottom body area. sensors will be mainly implemented in these areas.

Among all geometric shapes, a dot (sphere) is the simplest and most symmetric shape. Dew is designed with a tweak on this shape, with its head-tip as a small and smooth extrusion, to break away from the perfect symmetry, and give a further trigger of sensation. The shape partially originated from an imaginary cartoon figure of a lively and amiable red-hat boy, who is always looking around at its surroundings. The tip of the shape has its inspiration from water droplets, plant buds as well as morning dew, objects that indicates liveliness, health, energy and happiness. The smooth and floating outlook welcomes people to touch and interact. It feels soft and tender, conveyed through both the form of its outlook as well as choices of materials. One proposal for the material is translucent low density polyethylene covered with translucent silicone, allowing light to pass through, also gives a tender and warm touch to the outer layer, allows it easy to hold on to.