LEARNING PROCESS
Establishing a communication channel between programming software and simulation environment.
​
This is me learning how to establish synchronized communication between programming and simulation environment. MATLAB is used to code and V-REP is used as a robotic simulation environment. Once the MATLAB code is executed, the motion of the robot can be seen in V-REP and then those images of the RGB-D camera sensor of the robot can be seen in the MATLAB. So, two-way-communication has been established.
LEARNING PROCESS
Turning the motors in a simulation environment
​
​This is me trying to replicate moving obstacles by giving wheels to the piece of metal and also learning how to turn those motor in a way I want.
LEARNING PROCESS
Image Processing and Obstacle Tracking
This is me trying a basic computer vision image processing technique. Here, I am tracking the object seen by the robot via MATLAB code. Image is converted to greyscale to save the computation time. Here I would like to mention that there are no computer vision toolbox/library is used. The real-time obstacle tracking is entirely hard-coded by me.
Implementation
This is me using the single obstacle tracking technique and implementing to track actual moving obstacles in the environment. This is the video I used as my mid-term presentation for the Probabilistic Robotics class.
LEARNING PROCESS
Video/Image processing and Multiple Obstacle Tracking
​
This is exactly the same computer vision technology as the previous video but with a small update. Here I am tracking multiple obstacles together and using their own color signature to differentiate with each other. Again the video is converted to greyscale in MATLAB to save computational time and the entire computer vision part is hard-coded by me.
Implementation
After learning how to keep track of multiple obstacles, it was implemented to moving obstacle. I used this video as my final presentation for the course. Code for the particle filter and how the motion of those obstacles can be predicted using that is explained in the report.
Just Playing Around!
Pose Estimation Using Kalman Filter
​
This is me just having fun with V-REP. Here I am trying to predict the pose of a robot based on the last measurement and using that measurement into the Kalman filter algorithm. In the plots given below the prediction without Kalman filter and with Kalman filter can be seen with the true pose. This is a good example to show what noisy data can do to our prediction if the filter is not applied.
