Computer Vision based robotic guidance system have been developed and deployed in this project. The project initiated with the literature review of the research papers in the field of robotic guidance. The information extracted have been used to develop an in-house Computer Vision (CV) based robotic guidance system. Python 3.5 programming language has been used with Spyder3 IDE to develop the algorithm of robotic guidance system, as well as to control the robotic hardware. Raspberry Pi 4B 4GB single board computer is used to execute the guidance system to control a robotic vehicle with the onboard camera feedback. V-rep robot simulation software is used to simulate the robotic guidance system algorithm in virtual robot before it is deployed to the physical robot. The virtual robot used is also developed in-house with Autodesk Inventor CAD software with all the parameters of the physical robot. The robotic guidance system performances are analysed, potential use of the robotic guidance system evaluated and documented. At the end, this report concluded with the future improvement options of the developed robotic guidance system.
Detail of my work can be accessed from www.indiceuk.com
3. CV Based Sensors
Provide superior feedback through image processing than the reflection-based sensors
Contains higher number of feedback information the image
Cheaper to procure
Easy to maintain due to a least number of non-contact parts
Low power consumption
4. Aim of the project
This project is aimed at developing an efficient, deployable Computer Vision (CV) based
Raspberry Pi 4 powered mobile autonomous robot vehicle.
5. Project Objectives & Process Chronology
Virtual Robot
Simulation test bed
P
I
D
controller
Raw
code
Commissioned
for
deployment
P
I
D
controller
Virtual Robot
6. Process Chronology
Virtual Robot
Simulation test bed
P
I
D
controller
Raw
code
Commissioned
for
deployment
P
I
D
controller
Virtual Robot
Raspberry Pi 4 with Heatsink Case
DRV84 Dual
Bridge motor
controller
Webcam
8. Process Chronology
Virtual Robot
Simulation test bed
P
I
D
controller
Raw
code
Virtual Robot
Commissioned
for
deployment
P
I
D
controller
Detection
Recognition
Global coordination
Distance measurement
22. Identifying wall and robot localization
Start
Find object
coordinate
Object
Location?
Location >1 and Left
Location >1 and Right
Location >1 and
Centre
Location <1
Left Pair Wheel
(Clockwise)
Right Pair Wheel
(Clockwise)
Both Pair Wheel
(Clockwise)
Left Pair Wheel (
Anti-Clockwise)
Program Terminated
Reset & End
23. Utilizing CNN for navigation
Convolutional Neural Network
(CNN) has been used to classify
object in target and obstacle
This information is used to provide
guidance to the robot to move
forward or navigate against
24. Project Outcome
The robotic guidance system developed is able to identify targets and navigate
to it
It can identify the obstacles (walls) and navigate against it
It can measure the target and obstacle distance, thus it can localize it’s own co-
ordinates
It follows a red blob and keep 1 feet distance away
It navigates against the wall
27. Project process reflection
Field of advanced robotic has been explored
Python robotic capability explored
Robotic hardware integration practiced
Real life entrepreneurship option explored
28. Project Conclusion
• The project aim and objectives have been accomplished.
• A computer vision based robotic guidance system have been designed and
deployed
• It’s performance have been analyzed and documented
• Future project works will be to make the operational parameters more precise
and multi programming environment deployable.
30. List of References
Bellis, M. (Jul. 3, 2019, thoughtco.com/definition-of-a-robot-1992364.) the
Definition of a Robot [online] available from <www.thoughtco.com/definition-of-a-
robot-1992364.> [11/10 2019]
Billingsley, J. e. and Brett, P. e. (2015) Machine Vision and Mechatronics in
Practice. 1st ed. 2015.. edn
Chaumette, F. (2015) Potential Problems of Unstability and Divergence in
Image-Based and Position-Based Visual Servoing.
Chesi, G. and Hung, Y. S. (2007) 'Global Path-Planning for Constrained and
Optimal Visual Servoing'. IEEE Transactions on Robotics 23 (5), 1050-1060
Dastur, J. and Khawaja, A. (2010) Robotic Arm Actuation with 7 DOF using Haar
Classifier Gesture Recognition.
Di Castro, M., Almagro, C. V., Lunghi, G., Marin, R., Ferre, M., and Masi, A.
(2018) Tracking-Based Depth Estimation of Metallic Pieces for Robotic Guidance.