Research and Documentation of My Work in Robotics and Computing

High-Density Tactile Sensing Artificial Fingertip for Humanoids and Prosthetics

This project focuses on the design and development of a high-density tactile fingertip module intended for humanoid robot and prosthetic hand applications. The module is capable of measuring pressure exerted on its surface through 22 individual pressure sensors distributed across the fingertip area. It is developed using Flexible Printed Circuit Board (Flex-PCB) technology.

In this project, an origami technique is employed to fold a flat, paper-like flexible tactile sensing module into a three-dimensional shape that mimics the form of a human fingertip. Sensor data acquisition is handled by an onboard 32-bit, low-power microcontroller, and the module communicates using the SPI protocol.

In this project, an origami technique is employed to fold a flat, paper-like flexible tactile sensing module into a three-dimensional shape that mimics the form of a human fingertip. Sensor data acquisition is handled by an onboard 32-bit, low-power microcontroller, and the module communicates using the SPI protocol.

The development of this module presented a number of challenges, particularly in mechanical design and electronics. Prior to this project, I had limited experience with mechanical Computer-Aided Design (CAD). I had to learn how to design both mechanical structures and electronic schematics, as well as manage the manufacturing and integration processes.

Despite these challenges, a functional prototype of the tactile fingertip was developed. While the current version is not yet optimized for production, it is capable of capturing tactile data. Ongoing work includes further optimization of the design, data collection, and analysis.

The dimensions of the fingertip module closely match those of a typical human fingertip, enabling applications in humanoid robots and prosthetic systems.

CAD Model

Figure 2: The CAD model on the left illustrates the structural design and assembly of the high-density tactile fingertip module. It was created using Siemens Solid Edge, leveraging the sheet metal environment to model the foldable geometry of the flexible PCB. The part environment was used to design and integrate additional components, including pressure sensors and the onboard microcontroller.

Board Layout

The schematics and board layout of the tactile fingertip module were designed using Autodesk Eagle EDA. It was a challenging task, as a large number of vias had to be managed within a very small area. The board layout includes approximately 218 vias on a two-layer flexible PCB, with dimensions around 32 mm × 74 mm.

Figure 3: Board layout of the tactile fingertip module.

Fully Populated Tactile Fingertip Module

The tactile fingertip module is manufactured using 2-layer flexible PCB technology with a thickness of 0.11 mm. After fabrication, surface-mount technology (SMT) is used to populate the electronic components. Figure 4 depicts the fully populated tactile fingertip flexible PCB placed next to a 10-cent coin, highlighting the size difference and the complexity of the design.

Figure 4: Fully fabricated and populated tactile fingertip module in flat form factor.

Origami

A technique called origami, a traditional Japanese art of folding paper to form intricate shapes, is applied here to transform the flat version of the flexible tactile fingertip module into a 3D shape resembling a fingertip. Although origami is a complex art and I don't have much expertise in it, I attempted a simplified version to apply the technique as best as I could. The figure below shows the fingertip after being folded using origami techniques, including both the top (Figure 5a) and bottom (Figure 5b) views. The bottom view represents the fleshy part of the fingertip.

Figure 5: Top (a) and Bottom (b) views of the tactile fingertip module folded into a 3D humanoid fingertip-like shape using origami techniques.

Tactile Fingertip on a Human Finger

Figure 6: Shows the high-density tactile fingertip module mounted on a human finger. This demonstration is for illustrative purposes only. While the module is designed for integration into a humanoid robot hand, I did not have access to the robot hand toward the end of the project. I plan to update this demonstration with a robotic hand implementation in the near future.

Future Work

Further work on data collection, analysis, and system optimization is planned. In future updates, I aim to include additional documentation such as:

Detailed experimental results
Data analysis and sensor performance evaluations

This project will likely continue to evolve as time allows. At the moment, my availability is limited due to involvement in other work. While I have clear plans for improvements, progress will depend on available time."

I haven't published the data, schematics, or full documentation yet, as I believe the project isn’t complete enough to share publicly. Once it reaches a point I consider satisfactory, I will make those materials available.

AI Traffic Monitoring Application with Object Detection and Counting

This AI application detects traffic objects, such as cars, trucks, bicycles, motorcycles, and pedestrians, using computer vision, and counts them as they cross a traffic signal. It uses a pretrained YOLO (You Only Look Once) model for object detection in images, video frames, or live camera feeds.

The application was tested on an NVIDIA Jetson Nano for live stream processing and on Ubuntu 24.04 for processing prerecorded videos.

This project can be extended to other use cases, including:

Parking lot occupancy detection
Pedestrian flow monitoring
Traffic anomaly detection (e.g., driving against traffic)
Real-time lane usage and congestion tracking

This version of the application is implemented in Python and does not operate in real-time. I have also implemented a C++ version of the application, which works in real-time. The C++ version will be published once the installer is ready, as its dependencies currently cause installation issues across different Ubuntu versions.

For more details, visit my GitHub page: https://github.com/neoviki/ai.vision.traffic.monitor.app

References:

[1] Ultralytics YOLOv8 Documentation: https://docs.ultralytics.com/models/yolov8/
[2] Ultralytics GitHub Repository: https://github.com/ultralytics/ultralytics
[3] Video by German Korb on Pexels: https://www.pexels.com/video/road-systems-in-montreal-canada-for-traffic-management-of-motor-vehicles-3727445/

Handwritten Digit Recognition Application using CNN

A simple and intuitive application that recognizes handwritten digits using a Convolutional Neural Network (CNN). The model is trained on the MNIST dataset and deployed with a user-friendly PyTkinter GUI.

Getting Started

To get started with the Handwritten Digit Recognition App, follow these steps:

Clone the repository
Navigate to the project directory
Install the required dependencies
Run the application

Clone the Repository


git clone https://github.com/neoviki/ai.vision.handwritten.digit.recognizer.cnn.mnist.git
cd ai.vision.handwritten.digit.recognizer.cnn.mnist

Install Dependencies


chmod +x install.dependencies.sh
./install.dependencies.sh

Run the Application


chmod +x run.app.sh
./run.app.sh

Reference


@article{lecun2010mnist,
  title={MNIST handwritten digit database},
  author={LeCun, Yann and Cortes, Corinna and Burges, CJ},
  journal={ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist},
  volume={2},
  year={2010}
}

For more details on this project, visit https://github.com/neoviki/ai.vision.handwritten.digit.recognizer.cnn.mnist

IoT Data Transmission Suite

An IoT Data Transmission Suite with a server daemon and client library, designed for IoT applications. Traffic is transmitted over the HTTP/HTTPS protocol. It supports homegrown plant monitoring and vehicle status tracking, offering a cost-effective alternative to MQTT for lightweight, low-traffic communication. This suite is ideal for simple IoT ecosystems where frequent or intensive data exchange is not required.

For more details on this project, visit https://github.com/neoviki/iot.data.exchange.suite

Tactile Controller Module

The Tactile Controller Module is an interface board designed to read tactile sensor data from a dedicated sensor module containing 36 sensors. It is powered by the PIC32MM0256GPM036 microcontroller in a UQFN 40-pin package. The module is built to sample and manage data from high-density tactile sensor arrays, offering flexible communication options via I²C, hardware SPI (HW SPI), and software SPI (SW SPI) protocols.

Features

Microcontroller: PIC32MM0256GPM036 (UQFN-40)
Protocols Supported:
- I²C
- Hardware SPI (HW SPI)
- Software SPI (SW SPI)
Application:
- Interfaces with high-density tactile sensor arrays
- Suitable for applications requiring customizable sampling and interfacing logic

Use Case

The controller module is primarily used to sample sensor data from a high-density tactile sensing module with 36 sensors. It communicates with the tactile sensing module using a software SPI protocol. Designed for applications such as robotics, touch-sensitive interfaces, and pressure-mapping systems, the tactile sensing module provides detailed tactile measurements, specifically pressure data. The controller runs essential firmware to acquire and format the sampled data, and transmits the tactile sensor data to an external system (such as a connected computer or robot) via a hardware SPI protocol.

Populated Module

For more details on this project, visit
https://github.com/neoviki/tactile.controller.module

Real-Time Traffic Monitoring Using Darknet (YOLO)

This project was developed for a client who required real-time counting of vehicles crossing a traffic signal. I used the YOLO (You Only Look Once) object detection algorithm, implemented through the Darknet framework. In the prototype version, the AI logic runs on an NVIDIA Jetson Nano.

The system processes video streams to detect and classify objects such as cars, pedestrians, and traffic signs, using a pretrained YOLO model based on the Microsoft COCO dataset. YOLO is a deep learning model that applies a convolutional neural network to perform object detection quickly and accurately in a single pass, making it well-suited for real-time applications. In this project, I used the model as-is for the car counting task, without training the network or refining it further.

Note: The demo shown above uses sample footage, not a live traffic feed, as I do not have access to the deployed device.

The vehicle counting functionality is proprietary and was delivered only to the client. However, the remaining implementation has been made public on GitHub. The published version is configured and tested on Ubuntu 16.04 using a ThinkPad T420.

This project can be extended to other use cases, including:

Parking lot occupancy detection
Pedestrian flow monitoring
CCTV-based object recognition for security
Traffic anomaly detection (e.g., wrong-way driving)
Real-time lane usage and congestion tracking

For more details, visit my GitHub page: https://github.com/neoviki/ai.vision.realtime.traffic.monitoring

References:
Darknet by Joseph Redmon : https://pjreddie.com/darknet/
Microsoft COCO Dataset : http://cocodataset.org/#home

Tactile Sensing Module with a Dense Arrangement of Sensors

This module features a high-density arrangement of 36 MEMS barometric pressure sensors in a compact 50x65 mm form factor, with potential applications in robotics and other fields. The goal is to explore the sampling frequency, potential hysteresis issues, and other influencing factors. The module utilizes 36 MEMS barometric pressure sensors. In addition to this hardware, a separate controller module (GitHub) has been developed to interface with the module and sample data.

At the moment, only part of the tactile sensing hardware is populated and tested, as I’ve been occupied with other projects. I plan to resume work and publish the complete version along with test results when I have more time.

Figure 1: Shows the partially populated tactile sensing module with MEMS barometric pressure sensors.

Tactile Sensing Module with 36 Sensors — **Figure 1**

Board Layout Top and Bottom Layer — **Figure 2:** Top (a) and bottom (b) layers of the tactile sensing module's board layout with 36 pressure sensors.

The schematics, evaluation results, and additional details will be shared once testing and evaluation are complete. I’m documenting the progress so far, but since this is a personal project, I need to prioritize other commitments for the time being.

Possible Applications:

Possible applications of this tactile sensing module include robotics, touchpads, gaming interfaces, and other fields that can benefit from tactile sensing.

List of Projects

I've had the opportunity to work on projects across a range of fields, including telecom, vehicle telematics, IoT, computer vision, and robotic sensor design - covering areas such as computer science, electronics, and robotics. These projects were practical solutions to challenges I encountered in both day-to-day computing and industry environments. They include Linux utilities, drivers that optimize and streamline personal and industrial workflows, Artificial Intelligence driven applications and command-line tools using `awk`, `sed`, and regex for data splitting, filtering, manipulation, and interpretation. I've also developed C/Python libraries and packages, as well as custom hardware modules such as tactile sensors for robotic sensing.

Below is a list of selected projects, all of which I've published on GitHub.

Nr	Project Details	Reference
1	ai.vision.realtime.traffic.monitoring	Link
2	flexible.tactile.sensing.module.dual.sensors	Link
3	tactile.controller.module	Link
4	ai.vision.handwritten.digit.recognizer.cnn.mnist	Link
5	ai.vision.image.classifier.cnn.cifar10	Link
6	ai.ml.classification.knn	Link
7	ai.vision.face.detector	Link
8	explainable.ai.evalution.app	Link
9	linux.over.the.air.app.updater	Link
10	linux.tcp.port.tunneler	Link
11	linux.remote.debugger.utility	Link
12	linux.app.installer.utility	Link
13	linux.embed.dir.to.bash.script	Link
14	linux.rpi.samba.file.share	Link
15	linux.rpi.vnc.server	Link
16	linux.rpi.wifi.access.point	Link
17	linux.wifi.monitor	Link
18	mesos.distributed.cluster.deployer	Link
19	proxmox.virtualization.cluster.deployer	Link
20	ceph.distributed.storage.cluster.deployer	Link
21	compress	Link
22	versionsnap	Link
23	graphical.interface.grid.eye.sensor.electron	Link
24	graphical.interface.gyroscope.sensor	Link
25	graphical.interface.robotic.gripper	Link
26	graphical.interface.serial.monitor	Link
27	iot.data.exchange.suite	Link
28	eagle.library.pic32mm0256gpm036.uqfn.40pin	Link
29	eagle.library.pic32mm0256gpm028.uqfn.28pin	Link
30	arduino.AMG8833.grid.eye.sensor	Link
31	arduino.Bosch.BMP384.pressure.sensor	Link
32	arduino.VL6180X.time.of.flight.distance.sensor	Link
33	arduino.esc.controller	Link
34	audio.joiner	Link
35	audio.multiplier	Link
36	video.splitter	Link
37	video.to.gif	Link
38	automate.transaction.on.ebay	Link
39	crop.image	Link
40	docker.wrapper.commands	Link
41	gps.coordinates.renderer.osm	Link
42	ipc.filelock	Link
43	ipc.messenger	Link
44	ipc.shared.memory	Link
45	mac.osx.tabrun	Link
46	netcopy	Link
47	rand.file.tagger	Link
48	terminate	Link
49	uscreen	Link
50	website.backup.and.restore.utility	Link
51	html.ui.generator	Link
52	ui.python.tkinter.wrapper	Link

Robotic Sensor Module Design

Tactile Sensing Artificial Fingertip for Prosthetics and Humanoids

AI - Computer Vision

AI Traffic Monitoring Application with Object Detection and Counting

AI - Computer Vision

Handwritten Digit Recognition App Using CNN

Computer Networks - IoT

IoT Data Transmission Suite

Embedded Controller Module

Tactile Controller

TUI - Terminal User Interface

Terminal CSV Editor

Sensor Module Design

Tactile Sensing Module with a Dense Arrangement of Sensors

Other Projects

List of Other Projects I have Worked On

About Me.

Contact Me.