Humanoid Robot Eye : Part 2

In this part of the series we will discuss some small but necessary terms related to humanoid robot eyes. There are some facts that are necessary to be understood while designing Humanoid robotic eyes. After all, it's all inspired by humans. So lets first discuss Saccadic motion of the Eye. It is a fact which everyone knows that a human eye can not be static. It always moves even if it is looking at a painting. His eyes always makes small movements.

Saccadic motion: A saccadic motion is a very fast movement or motion of the eye. It is actually a fast shift of the frequency of the emitted signal.

 

Recording of Horizontal saccades

Saccades are the fastest movements produced by the human eye. They take about 210 miliseconds to be initiated. During a Saccade, the peak angular speed of the eye reaches upto 1000/sec. These saccades are generated by a neuronal mechanism. This mechanism activates the eye muscles more directly.

The amplitude explains how much an eye has covered in the angular distance during any movement. Untill an eye makes an angle of 60, the velocity of a saccade linearly depends on the amplitude. But for saccades whose angular distance measurement is larger than 60, the peak velocity starts varying non-linearly.

These saccades are classified into four parts: Visually guided saccades, Anti-saccades, Memory-guided saccades and predictive saccades. Memory guided saccades are very accurate in measurement.

Visually guided saccades: Visually guided saccades are the saccades which are sensitive for any change visually. This type of saccades are further classified into two categories: Reflexive and scanning. The reflexive saccades are triggered exogenously, whereas scanning saccades are done endogenously. It is to be noted that scanning saccades are triggered in the way to explore the visual environment.

In an antisaccade, the eye has nothing to do with the visual things. The eyes move away from the visual onset. In this type of saccades the observer is not guaranteed about what he is watching and he makes many errors.

In memory guided saccades, the eyes make movement towards objects for which some memories already exists related to those objects. In this case no any additional visual stimulus is required.

There is an another type of saccades which is generally known as predictive saccades. In predictive saccades, the eyes are kept on an object which is moving in a temporary manner. These are some basic things which, although we will not use directly in designing a Robot eye, but as a basic knowledge it is necesarry that we all will be familiar with it.
                                               

Another thing which is very necessary to be familiar with is saccadic masking. It is a kind of phenomenon where the mind blocks visual processing during the movements of eyes. In the case of saccades, it is very common that the retina of a human eye gets blurred images as well because the retina of the eye possibly sweeps some images, the images which are blurred are cut by a inner generated process. During the phenomenon, the working of eyes almost become ineffective and its capacity becomes nil, equivalent to an eye of a blind man. But for a normal eye its duration is very small, nearly in microseconds. This phenomenon is called saccadic masking.

We will discuss the design of a Robot eye. To design any robot eye, the first requirement is the eye-ball. We need to design the spherical shape structure of the eye-ball.

The next thing we need is the actuation system which is necessary to provide ocular motions. The ocular motion is related to the pupil, eye-movements, eye-lids and lens. The saccadic motion we have discussed above is a part of eye-movement.

Now the thing to be seen is how we provide support to the eye-ball. The actuation and control system with the additional control system is also needed in the working of these robotic eyes.

The Robot eye-ball should be capable of hosting a commercial CMOS camera, video signals cables to external electronics. The eyeball is comparatively larger than the human eye, it should not be the question why it has to be because in making an eye ball we need lots of equipments as well some of them we have discussed above the remaining I am going to.

A very important thing is the supporting system without which we can not imagine a stable robotic eye. The supporting structure is the structure which is designed just to support the eye ball.

It has two parts; one is to hold the eye-ball and the other part is to implement the pointed pulleys. The second part which is used for the pulleys is called a Flange. The whole supporting system has to follow the Listing Law. The eyeball support is structured as two C-shaped parts which are joined together either by welding or by using any other similar phenomena. Its back part is meant for passing of video signals and power supply but actual contacts are on the C-shaped parts which make their connection with the eyeball.

One of the two parts which we have discussed before, the Flange, is used to implement the point wise pulleys.

The pulley is made tangent to the principle direction at any point. Suppose there is another listing compatible position. If the pulley tilts about the principle axis,the point of tangent wil be the same.

This is to allow the tendon to rotate through the tangential point in that position also. Now these robot eyes works on sensing and actuation. For that we need four DC servo motors which produce a maximum output torque in the range of several nanometers. These DC motors pull four independent tendons.

The servo motors we use are equipped with optical encoders which provides the main feedback for controlling ocular movements. Now the second set of sensors for the measurement of mechanical tension of the tendons on their excessive loads are avoided.

These sensors are actually Infrared based which is separated by a mobile shutter. The very essential part we will discuss is the Control system.

It has two levels. The lower level has two control loops for each actuator. The tension feedback control loop makes the eyeball back-drivable. This is done so that we can position the eye by hand as well. There are two types of control type actions. The P type control type action and P-I type control action. The P-type control action regulates the tension of the tendon while the P-I type control action is there to control the motor velocity.

The high level position based control loop is implemented by a PC based computer and the low level control algorithm is implemented on a multicolor board. We also use slave microcontrollers; one microcontroller is used for each motor.

These microcontrollers operate in parallel and are coordinated by a master one managing the communications through the Bus, either with sensing modules or with high-level control modules. At last it is also kept in mind that master and slave microcontrollers communicates using a multiplexed high speed.

Resources

Humanoid Robot eye-Part1

How to make a Robot-Part2