Academic Projects
Mastering the art of mechatronics & robotics, one project at a time…
Tele-Operated Dental Surgical Robot
I realised that the field of robotic assistive surgery is growing faster than ever. Companies like Intuitive Surgical have demonstrated with their daVinci Surgical System how robots can improve the accuracy and reliability of surgical procedures by assisting surgeons intra-operatively.
This was a research-driven project. Alongside 2 colleagues, we wanted to find a gap in the assistive surgical robot market. We wanted to find a surgical procedure for which no existing system can assist with. The procedure had to be relevant and timely. After a rigorous literature review, we decided to conceptualise a tele-operated surgical system that assists with maxillofacial surgeries, especially with impacted wisdom teeth extraction.
Impacted (buried) wisdom teeth are usually very close to both the lingual and inferior alveolar nerves. These are statistically susceptible to damage by surgeons during such surgeries. Our system allows surgeons to remotely control the surgical device that has all the tools required. It uses 3D frame mapping to find the relative position of these 2 sensitive nerves and makes it impossible for the tool to hit them.
We designed the whole system, including the required customised tools, as well as its control system.
This was a research-driven project. Alongside 2 colleagues, we wanted to find a gap in the assistive surgical robot market. We wanted to find a surgical procedure for which no existing system can assist with. The procedure had to be relevant and timely. After a rigorous literature review, we decided to conceptualise a tele-operated surgical system that assists with maxillofacial surgeries, especially with impacted wisdom teeth extraction.
Impacted (buried) wisdom teeth are usually very close to both the lingual and inferior alveolar nerves. These are statistically susceptible to damage by surgeons during such surgeries. Our system allows surgeons to remotely control the surgical device that has all the tools required. It uses 3D frame mapping to find the relative position of these 2 sensitive nerves and makes it impossible for the tool to hit them.
We designed the whole system, including the required customised tools, as well as its control system.
Autonomous Firefighting Drone
Working with a few colleagues, we took on a challenge to create an end-to-end concept for an autonomous firefighting drone.
We went about the project from a very practical perspective as we wanted to create a concept that is feasible and that could actually be adopted by emergency services to assist firefighters. We put real-world limitations such as a maximum frame size and a weight limit (including the weight of the water). We also put a minimum operational flight time and a minimum thrust-weight ratio.
It was challenging to incorporate all these limitations. The whole design process and component selection process involved many trade-offs. We had to learn much about the science behind drone flights and navigation algorithms using LIDAR sensors.
We managed to finalise a hexacopter concept that fulfilled these criteria:
- Autonomous indoor/outdoor navigation
- Able to search unfamiliar indoor spaces with low visibility
- Detect fires and autonomously navigate towards them
- Automatically spray suppressant liquid onto the fire source
- Has a minimum thrust-weight ratio of 1.8.
- Carries 4 liters of suppressant liquid.
We went about the project from a very practical perspective as we wanted to create a concept that is feasible and that could actually be adopted by emergency services to assist firefighters. We put real-world limitations such as a maximum frame size and a weight limit (including the weight of the water). We also put a minimum operational flight time and a minimum thrust-weight ratio.
It was challenging to incorporate all these limitations. The whole design process and component selection process involved many trade-offs. We had to learn much about the science behind drone flights and navigation algorithms using LIDAR sensors.
We managed to finalise a hexacopter concept that fulfilled these criteria:
- Autonomous indoor/outdoor navigation
- Able to search unfamiliar indoor spaces with low visibility
- Detect fires and autonomously navigate towards them
- Automatically spray suppressant liquid onto the fire source
- Has a minimum thrust-weight ratio of 1.8.
- Carries 4 liters of suppressant liquid.
Prosthetic Limb Intent Predictor (Neural Network & SVM)
With the increasing number of people dependant on prosthetic limbs, lots of research has gone into the design of much stronger and more reliable artificial limbs. However, there are still gaps in reliable methods of controlling them. The device (prosthetic or exoskeleton) needs to know what movement the user is intending to be able to assist them with it. I decided to contribute to this field.
Training data was collected from healthy volunteers wearing 7 IMU sensors as they performed 5 activities of daily living, such as ascending stairs. I extracted 5 time domain features and then used this data to find the sensor location providing the most relevant data for classification (using Fast Fourier Transformation). I then built a pattern recognition Neural Network as well as a multiclass Support Vector Machine model and trained them both with that data.
After fine-tuning model parameters, I managed to successfully predict the user intended activity with an accuracy of over 96% using a single IMU sensor.
The project is now open-sourced.
Training data was collected from healthy volunteers wearing 7 IMU sensors as they performed 5 activities of daily living, such as ascending stairs. I extracted 5 time domain features and then used this data to find the sensor location providing the most relevant data for classification (using Fast Fourier Transformation). I then built a pattern recognition Neural Network as well as a multiclass Support Vector Machine model and trained them both with that data.
After fine-tuning model parameters, I managed to successfully predict the user intended activity with an accuracy of over 96% using a single IMU sensor.
The project is now open-sourced.
A Robotic Arm Controlled by Neuromuscular Interface
This was my third year engineering bachelors project.
In a nutshell, I worked on using EMG sensors to detect neuromuscular signals from the upper limb to control a robotic arm. I implemented machine learning algorithms to train my system for accurate gesture recognition.
Stroke survivors suffer from severe muscle weakness, especially in their arms. My research contributed to the massive study of neuromuscular rehabilitation, bringing such survivors one step closer to being able to control prosthetic hands. I demonstrated, for the first time, the feasibility of using 3D-printed electrodes in capturing neuromuscular data non-invasively.
Innovating a way to advance in this field led to an invitation to present my research paper at the International IEEE/ASME Advanced Intelligent Mechatronics conference (Hong Kong, 2019). This was truly an incredible experience.
The hardware was modeled with the patient's convenience in mind. I created a wireless, wearable device that can easily be used:
In a nutshell, I worked on using EMG sensors to detect neuromuscular signals from the upper limb to control a robotic arm. I implemented machine learning algorithms to train my system for accurate gesture recognition.
Stroke survivors suffer from severe muscle weakness, especially in their arms. My research contributed to the massive study of neuromuscular rehabilitation, bringing such survivors one step closer to being able to control prosthetic hands. I demonstrated, for the first time, the feasibility of using 3D-printed electrodes in capturing neuromuscular data non-invasively.
Innovating a way to advance in this field led to an invitation to present my research paper at the International IEEE/ASME Advanced Intelligent Mechatronics conference (Hong Kong, 2019). This was truly an incredible experience.
The hardware was modeled with the patient's convenience in mind. I created a wireless, wearable device that can easily be used:
Robotic Colonoscopy Arm
Early detection is paramount with cancer. Colonoscopy is the most effective method for diagnosis. Patients, however, fear the current colonoscopy method due to its invasiveness, discomfort and pain.
Researchers have been looking a different methods of undertaking colonoscopies using freely moving capsule robots. Since locomotion in such an environment is challenging, a solution is the use of magnetic capsules that can be guided around the colon using an external magnet mounted on a robotic arm.
This leads to a less invasive method of undertaking colonoscopies which in turn may lead to an increased uptake in screening for CRC in the UK and worldwide.
I worked in a team to design an autonomous mechatronic arm using SolidWorks and build it from scratch.
Its function was to guide a micro robot inside the body from the outside, hopefully revolutionising colonoscopy procedures by making them non-invasive, more accurate, and much easier to perform.
Researchers have been looking a different methods of undertaking colonoscopies using freely moving capsule robots. Since locomotion in such an environment is challenging, a solution is the use of magnetic capsules that can be guided around the colon using an external magnet mounted on a robotic arm.
This leads to a less invasive method of undertaking colonoscopies which in turn may lead to an increased uptake in screening for CRC in the UK and worldwide.
I worked in a team to design an autonomous mechatronic arm using SolidWorks and build it from scratch.
Its function was to guide a micro robot inside the body from the outside, hopefully revolutionising colonoscopy procedures by making them non-invasive, more accurate, and much easier to perform.
Autonomous Rover
A rover, with a decent suspension system, was to be designed and built. It had to be autonomous so it maintains a straight path regardless of how bumpy the terrain is. All that had to be done while making sure the rover drives as fast as possible.
The idea was to integrate negative feedback control to make sure the rover does not deviate from its path. This was meticulously implemented using labVIEW on cRIO.
As for the hardware, we designed a unique suspension system based on what we were taught on vibrations & control. The chassis as well was designed from scratch with durability, lightweight and compactness in mind. This was done on Solidworks.
This project was sponsored by Jaguar.
The idea was to integrate negative feedback control to make sure the rover does not deviate from its path. This was meticulously implemented using labVIEW on cRIO.
As for the hardware, we designed a unique suspension system based on what we were taught on vibrations & control. The chassis as well was designed from scratch with durability, lightweight and compactness in mind. This was done on Solidworks.
This project was sponsored by Jaguar.
An Autonomous Turtlebot Explorer
Within a group, I implemented a program that plays Robotic Cluedo.
Slightly different than the board game, Robotic Cluedo involves programming a Turtlebot that needs to autonomously visit a room in an arena and identify the character and weapon in the room along with their location.
Coding was done in Python. We used various computer vision algorithms along with intelligent path planning techniques for the bot to visualise its surroundings and explore on its own.
Slightly different than the board game, Robotic Cluedo involves programming a Turtlebot that needs to autonomously visit a room in an arena and identify the character and weapon in the room along with their location.
Coding was done in Python. We used various computer vision algorithms along with intelligent path planning techniques for the bot to visualise its surroundings and explore on its own.
C++ Breakout Game
I put together and soldered a game controller with buttons, potentiometers and other elements, along with an mbed microcontroller.
I then coded onto it a unique version of the game Atari Breakout using C++ language to be played using the buttons and various controllers on the gamepad.
I then coded onto it a unique version of the game Atari Breakout using C++ language to be played using the buttons and various controllers on the gamepad.
Remote-Control Buggy
As a team, we built a remotely controlled car which functions on high- frequency radio waves. My role was ensuring the flawless functionality of the wireless transmission of signals, involving encoders, decoders, transmitters and receivers. I also programmed an Arduino Uno to implement control of the buggy using an analogue joystick.
