Embedded Systems - Indian Institute of Technology Kanpurstudents.iitk.ac.in/roboclub/lectures/Embedded Systems.pdf · embedded systems indian institute of technology kanpur 4th inter

Embedded SystemsINDIAN INSTITUTE OF TECHNOLOGY KANPUR

4TH INTER IIT TECHNICAL MEET

Objective

Our Objective was to design and fabricate a voice controlled robot integrated with image processing abilities. The robot maneuvers by identifying voice commands and semi-autonomously detects and picks certain objects using image processing.

VAVER v1.2Voice Automated and Vision Empowered Rover

We started working in mainly 3 departments required for VAVER i.e. Design and fabrication, Image Processing and Voice Recognition.

As our objective was to maneuver the robot in many directions, so we decided to use omni-wheels.

Also, as we have to keep our robot’s gripper always aligned to the object, we decided to put the camera just above gripper and Raspberry-Pi as the processing required was not too heavy.

For Voice recognition, we used the very versatile and easily accessible software MATLAB and MFCC library.

VAVERVAVER

DesignDesign Image Processing

Image Processing

Voice Recognition

Voice Recognition

DesignThe problem statement required the bot to maneuver in all directions while being controlled by voice and image data.

The rover can traverse in multi-directions and also has a gripper installed to grip objects.

So, the rover was designed as a 3 wheel drive using omni-wheels for multi-directional movement.

The chassis is triangular, with welded aluminum pipes, giving maximum strength while keeping the weight minimum.

Electronics and processors were efficiently placed to balance the weight and also keeping aesthetics in mind.

The gripper and camera were center aligned and were placed low so that objects won’t topple.

Gripper

We used four-bar mechanism for the gripping mechanism. Planar four-bar linkages are constructed from four links connected in a loop by four one degree of freedom joints. The joint is a revolute, that is a hinged joint.

We then used a servo to rotate the mechanized gear between the two bars for exact gripping and locking.

Image Processing

It can identify and isolate objects based on color, shape and size. It can also simultaneously track the objects giving the center and the size of the objects. We used the ramer-douglas-peucker algorithm for detecting the shape of the object. Also, it has integrated advance PID controls with object tracking to calibrate and set the range for gripping mechanism, resulting in significant increase in accuracy.

VAVER 1.2 has implemented the basics of OpenCV library and PID algorithm for real time tracking of objects. It has two tier system for tracking.

First it isolates the object from its surroundings based on the shape and the color of the object.

Then it forms a virtual contour for the identified Object and calculate its area using ramer-douglas-peucker algorithm.

Using certain limits for area of the contour and upon calibration, VAVER can pick the object on user command.

It aligns the object position with the gripper position using the x and y coordinates of the center of the object contour.

Voice RecognitionThe motion of VAVER 1.2 can be controlled using voice commands given by selected users. It can be controlled wirelessly within the limitation of network range. Users can be changed or authorized for VAVER once their database is ready and stored in the main database. Being a simple GUI based application, it is very User-friendly.

VAVER 1.2 has implemented well known MFCC and DTW method to command the rover using speech signals.

After receiving the voice command, it filters the noise from the input signal based on frequency and amplitude filters.

Then it computes the Mel Frequency Cepstrum Coefficients (indication of basic characteristics of voice).

It then shifts to “Dynamic Time Warping” method to compare the extracted coefficients with the database of coefficients of stored commands and identifies the one it is closest to.

Then it uses Raspberry pi Support package in MATLAB which uses TCP/IP connection system to transfer the command to R-pi which further sends it to Arduino board.

IntegrationIt has a single board computer Raspberry-Pi for establishing connection with the user. Motion voice commands are directed towards a microprocessor Arduino Uno board while tracking commands are directed towards R-Pi which then initiates the image processing capability of itself.

It uses TCP/IP connection to transfer the commands from MATLAB to Raspberry –Pi.

Being a single board computer, R-pi have system commands which are used to transfer voice commands to Arduino board and are also used for starting the object tracking capabilities of rover.

Arduino being a microprocessor further controls the motor and then the rover.

Innovation

Rover can move in 4 directions simultaneously. Also, we have designed our own MATLAB database and code which is specific for VAVER thus minimizing the chances of error in Voice recognition. We have also introduced a color and shape variable in our Image-Processing.

Multi-Dimension motion via omni-wheels

Own database and code of MATLAB minimizing risk of errors in Voice recoginition

Variable color and shape factors in Image Processing, thus making it more user-friendly as user can easily set the object.

Utility

Considering the present situations in world, VAVER 1.2 can be used in many field where a risk to human life exists. Detecting Bombs, wireless surveillance in inaccessible areas and also in fields related to day-to-day life, for ex: delivering specific items, helping in daily chores.

Detecting and carrying the bombs to a safe position.

Full controlled Wireless Surveillance in inaccessible areas.

Helping in day-to-day hand chores being a very cheap product.

Delivering specific items for some services.

Future Versions

In further versions we can introduce a database for image processing which can be used to detect multiple objects at a time. We can also add up new commands for the rover which can make it more flexible and more dynamic.

Introducing an app for full control of VAVER 1.2

Introducing a database for Image Processing codes which can be used to detect multiple objects at same time.

Adding more commands for VAVER via voice commands, improving its control ability and making it a multiple task performing rover.

Also, we can improve to a smaller, 8 directional rover using mecanum wheels.

Indian Institute of Technology Kanpur

Thank You