Friday, 12 February 2016

Team Pi Over Four

Sprint 1 Refection 

Author: Philip O Connor


Personal Reflection:
The goal set out at the beginning of this sprint stated that three key sections of the lampbotics design was required to be completed which were the face detection, servo control and sound control and was to be implemented using threads. The sections of the design I was involved with help from members of my team was the threads, a simple implementation of the servo control and the face detection section of the design. Though the participation of this group project allowed me to work on my communication skills though group discussions and problem solving sections of our lampbotics project design which are covered in the technical section.

Teamwork Reflection: 
At the start of the sprint the class was split up into different teams and included in my team was David O Mahony, Filipe Terra and Siang Wei Lee. The first team meeting assigned the different sections of the design of the lampbotics project which was:
  • Face detection was assigned to me 
  • Servo control was assigned to Filipe Terra
  • Sound control was assigned to David O Mahony
  • Threads was assigned to Siang Wei Lee
Learning outcomes obtained from this sprint demonstrated the need for duplicate components so if a member of the team is not present work on the project can precede. Also observed from this sprint was the need for more group testing interactions so that unforeseen problems could be solved which would allow more time to test the overall design. Also from not only team discussions but class discussions each of the different teams could pool their resources for the same sections of the design which would allow members of each team to cover a larger range of how that section could be implemented which would aid in the design of a more comprehensive singular design.
Project Management Reflection:  

The implementation for the first sprint allocated each of our members a different section of the design to work on. Once each of the members of our team had gotten a basic working model of their section the next section combined these sections together into one design. The singular design used threads to control different aspects of the design. To get a basic working model a simple threading design was used to implement the control of the different sections and preceding forward the thread design flow currently used will need to be altered to achieve faster processing of the different threads. Also Filipe Terra servo control work on the movement section of the design using the XMOS board to control multiple servos will greatly help with control of the various servos used in the lampbotics design.  
Furthermore observations made from the project management meets demonstrated the need for more group meets so that the different sections of the design could be interfaced together.  

Tech Reflection:  
  • Problems
The video uploaded for Team blog entire 2 demonstrate the simple implementation of the three key sections of the design for sprint 1. This demonstrated the operation of our design only using one servo and would be replaced with Filipe Terra servo control. Encounter problems from this test demonstrated the need for a more comprehensive system to process the threads due to the fact that the interacts between the face detection thread and the sound and servo control threads produced erratic behaviour which would cause the detected face to be lost. Another issue which was obtained from testing was that because of the fact that the current face detection module is based on detecting a persons front facial features which means that the detection of a persons face can be difficult. Furthermore in the test of the control of the single servo the position control of the servo large servo position changes which means that the detected face would be loss.
  • Solutions
A possible design solution for the processing of the threads would first require that mutexs be implemented to avoid race conditions which could cause unforeseen behaviour from the system. Also multi process could be used so that threads could be run in parallel which would help with eliminating errors caused from the time required to switch between threads in the current working model. Also to solve the position changes of the servos a possible design solution proposed by David which is shown in Team blog entire 2 and this could be used to solve the issue of large servo position changes. The issue with the detection of a persons face could be solved though two approaches which are:
  • Using trained cascades to detect a persons face
  • Or using an object which is more easily tracked to first obtain an optimal design to control the key sections of the design and then implemented the face detection
These two options could be used to solve the issue with the tracking and detection of a persons face.
  •  Ideas for next sprint
 I think for the next sprint the the lamp design should be examined because of the fact that current areas of the design may require a particular set up for example the positioning of the camera which will determine how the servos should be controlled.  
Another idea for further sprint work on the lampbotics project would be the transmission of the X, Y and Z position values to the internet wirelessly which would allow different lampbots to act as a hive.
Also to give more functionality to the lampbotics project the idea of introducing an actual light into the design would be interesting.  
Team Pi Over Four
Philip O Connor entry 4, 

No comments:

Post a Comment