0LAUK0 2018Q1 Group 2 - Design Plans Research

From Control Systems Technology Group
Revision as of 16:19, 23 September 2018 by S141153 (talk | contribs) (→‎Sources)
Jump to navigation Jump to search

Introduction

This page contains research performed by group 2 for the course Project Robots Everywhere (0LAUK0) regarding the state-of-the-art on actuators and face tracking. This literature study is divided into the following relevant topics:

  • face tracking
  • monitor-arm actuation


Face Tracking

Initiating the literature study

Wikpedia was used to get a quick glance of the idiom used when describing face tracking, as well as core information that helps informing the rest of the literature search.

The face tracking system that we intend to build should be minimally invasive. Given that this technology should be used comfortably during office work, we cannot force the user to wear special markers in order for the system to properly track them. This means that we will have to use a markerless (Wikipedia,n.d.) system, where facial features of the user (instead of dots or stickers placed on the user) have to be identified in the video input.

One of the requirements of our system is that it maintains an optimal viewing distance between the monitor and the user’s face. As such the face tracking system needs to be able to track depth. This means that we will need to implement three-dimensional tracking. This requires a multi-camera system (Wikipedia, n.d.)

Face recognition

Facial recognition is a technology using image analysis and computer vision to identify or verify people. Examples of facial recognition being used in practice are at airports, various surveillance systems etc. where facial recognition is used to verify the person’s identity and watch out for threats. Another example of facial recognition is iFace, the new facial recognition system build in the new iPhone X.

Why use facial recognition as a form of authentication [1]

A traditional password is often not the best choice for quite a number of people. Firstly, the strength of a password varies from person to person, about 14% of the people use a weak password for their email account. Secondly, more than two thirds of the people would fall for social engineering methods of hackers to obtain their passwords. Thirdly, around one sixth of the people would write their password down to remember it and around one sixth of the people would let at least one other person know their password. By using facial recognition as authentication, they don’t need to worry about creating strong passwords and remembering them.

Face recognition techniques [2]

There are two types of face recognition from intensity images, feature-based and holistic. Feature-based recognition approaches the problem by identifying and extracting prominent facial features of the person in the image, this is done via facial points and geometric relationships between the features. Afterwards an algorithm tries to match faces in the database with the extracted information of the face from the image. The advantage of feature-based recognition is that the size, orientation and lighting of the image are less relevant to matching the face in the image with one in the database. Making these factors less of an issue for getting a match. The disadvantage of feature-based recognition is the difficulty in choosing which of the facial features are actually important to matching the picture with the face in the database. Holistic recognition approaches the problem not only based on the face, but also based on the whole image.

Face recognition from Video Sequences [2]

This is real-time facial recognition based on actual image sequences, instead of a single image. Face recognition via video needs three components: A component that detects the face, a component that tracks the face and a component to recognise the face. To identify the people in the video sequence, most systems simply take a few good frames and apply face recognition techniques used for intensity images.

Face recognition using 3D modelling [2]

This method uses 3D modelling to get an accurate construction of the face using various sensors. By using this method, lighting, orientation and background clutter are of no issue when constructing the image of the face. [3]

Sources

  1. Šolić, K., Očevčić, H., & Blažević, D. (2015). Survey on Password Quality and Confidentiality. Automatika ‒ Journal for Control, Measurement, Electronics, Computing and Communications, 56(1). doi:10.7305/automatika.2015.04.587
  2. Jafri, R., & Arabnia, H. R. (2009). A Survey of Face Recognition Techniques. Journal of Information Processing Systems, 5(2), 41-68. doi:10.3745/jips.2009.5.2.041
  3. Retrieved from [1]
  4. Wikipedia. (n.d.). Facial motion capture. Retrieved from [2]

Monitor-Arm Actuation

To design the monitor arm we first needed to consider what actuation type to use. This actuation type is dependent on a few variables such as:

  • Precision
  • Torque
  • Speed
  • Noise
  • Cost
  • Design and looks
  • Compatibility
  • Input and output

Basically there are four convenient options for actuation. These are: Stepper Motors, Servo Motors, Direct DC Motors and Hydraulic Actuation. In the following paragraph we will explain what the pros and cons are of these actuation types based on the variables described above.

Stepper Motors

The principle of magnetic attraction and repulsion are used by stepper motors to move a screw. The torque is created by alternately applying current to the individual windings in the motor. This torque turns a permanent magnet rotor. Once the windings are energised, a holding torque is generated. Unlike DC (servo) motors, stepper motors have an inherent holding, detent or torque that can be used to maintain the position of devices in the power off state for a period of time. Steppers provide an inexpensive open-loop method to achieve high accuracy depending on the resolution. Furthermore, no feedback is required for position or speed control. Stepper motors are also compatible with modern digital equipment. They are also less expensive than servo motors, because of the removal of position feedback. On the other hand, steppers are bulky, noisy, generate a much amount of unwanted heat and do not provide smooth continuous motion. The large size of these motors can make it difficult to incorporate them into small mechanisms. Furthermore, friction can be a problem in the high accuracy control.

Continuous Rotation Servo Motors

When used with an encoder, servo motors provide smooth, continuous motion as well as high speeds and submicron accuracy. It consists of coils of wire around a metal core (called an armature) inside a magnetic field. When electric current is applied to the windings, the armature interacts with the magnetic field causing the armature to turn. These motors require constant power to maintain position, so it is not an ideal solution for set-and-hold applications. It also generates unwanted heat (not as much heat however than stepper motors) and requires a feedback mechanism for controlling position and velocity. Even when holding a specific position, these motors often oscillate around the position. In general, servo motors are considered to be more accurate than stepper motors, but just like with the stepper motors, friction can also be a problem in the high accuracy control.

Direct DC Motors

Direct DC motors are only usable when implemented in a way comparable to servo’s. The rotational speed of a direct DC motor is really high, and thus always needs some sort of gearing which makes a lot of nice and can be hard to implement. Direct DC motors also need an encoder to control properly. By doing this one actually makes a servo of the direct DC motor, so the rest is comparable.

Hydraulic Actuation

Hydraulic actuation can be a real silent type of actuation. This is only true when the pressurizer is not close to the monitor arm system. When controlled correctly it can be fairly precise, but it will always need some form of closed-loop feedback system. It may be difficult to find a sensor that is precise enough to use for this closed-loop feedback system. One could think of a linear solenoid actuator with some sort of encoder to drive the hydraulics, but this may make the system difficult to employ. Another problem with hydraulics is that it is based on a lot of assumptions on the flowing medium. This means that the output is always a bit different than the expected results. Hydraulics are a real powerhouse and always make sure the monitor arm stays in the right position as long as there are no leaks. Hydraulics are good at linear and rotary actuations. The speed of hydraulic actuation is not high, but probably fast enough for our system. Hydraulic systems are expensive compared to other actuation types. A hydraulic system can be implemented real neat when done right, but due to a lot of components it may look real messy and can give a lot of problems.

Chosen Actuation Type

We have chosen to use stepper motors for our robotic monitor arm. The stepper motor can precisely rotate to each wanted position for the robot arm. It is a good solution for the radial actuations needed. When implemented with the ‘Gispen duo zitsta desks’, which will be used in Atlas, the linear movement in the z direction of the monitor can be actuated by the same actuation used in the desk. This means that for the monitor arm only rotational actuation is needed. The biggest downside of the stepper motor is the noise it makes, but this can be solved by implementing stealthChop™ technology by Trinamic. When using a TMC controller device which uses stealthChop™ and spreadCycle™ the stepper motors noise can be reduced with more than 10 dB which makes the whole system silent enough to use in a workplace. When using a standard stepper motor such as the Nema 14, precision up to 1.8 degrees can be reached which makes the stepper motor accurate enough for the monitor arm. When the torque is not sufficient enough we could implement a spring system in the arms which reduces the force need to move the monitor arm but also makes sure the monitor arm stays in the right position without needing to much extra force from the stepper motor. The speed of the stepper motors is more than sufficient. SpreadCycle™ technology makes sure the stepper motor always operates smoothly. The costs for the actuation are not high compared to other solutions. Stepper motors are quite bulky, but should be easy compatible in a system. When integrated neatly, the motors should look quite nice.

Sources

  1. Ouyang, P.R., Tjiptoprodjo, R.C., Zhang, W.J. & Yang, G.S. Micro-Motion Devices Technology: The State of Arts Review. 2008. Retrieved from: [3]
  2. Tsui, K.W., Cheung, N.C. & Yuen, K.C. Novel Modeling and Damping Technique for Hybrid Stepper Motor. 2008. Retrieved from: [4]
  3. Gispen. ™ Duo Zitsta Bureau. 2010. Retrieved from: [5]
  4. Trinamic. StealthChop ™. 2017. Retrieved from: [6]
  5. Shanghai Moons Electric Company. Step Motors. Retrieved from: [7]
  6. DigiKey Electronics. TMC Silent Stepstick. Retrieved from: [8]
  7. Li, Y.F. & Wikander, J. Model Reference Discrete-Time Sliding Mode Control of Linear Motor Precision * Servo Systems. 2004. Retrieved from: [9]
  8. Techno Inc. Choosing Between Stepper and Servo Motors. Retrieved from: [10]
  9. Engineering 360 powered by IEEE GlobalSpec. DC Servomotors Information. Retrieved from: [11]
  10. Engineering 360 powered by IEEE GlobalSpec. Stepper Motors (Rotary) Information. Retrieved from: [12]
  11. Omega Engineering. Stepper Motors. Retrieved from: [13]
  12. Solarbotics. Stepper Motor Basics. Retrieved from: [14]
  13. Maske, R.J. & Woods, D.M. Use of Digital Current Ramping to Reduce Audible Noise in Stepper Motor. 2001. Retrieved from: [15]
  14. Gill, H. Stepper Motor or Servo Motor: Which Should It Be? 2016. Retrieved from: [16]
  15. Lackey, B. Servo Motor vs Stepper Motor: Which Is Right for Your Application? 2017. Retrieved from: [17]
  16. Erlich M. Steppers Versus Servos. 2017. Retrieved from: [18] (This link only provides part of the article. One needs to download the article in order to view it as a whole; it is not available online)