Hi, I'm Viisnu Anand S

This project was designed for the Smart India Hackathon 2024 (Software Edition) under the problem statement on enhancing wireless communication efficiency. Traditional systems use fixed modulation schemes, leading to inefficiency and increased error rates under changing signal conditions. Our proposed solution implements adaptive modulation and automatic modulation recognition (AMR) using deep learning integrated into a Software Defined Radio (SDR) testbed.

The system dynamically classifies modulation types — BPSK, QPSK, 16-QAM, 64-QAM — using real-time Signal-to-Noise Ratio (SNR) and Error Vector Magnitude (EVM) metrics. A custom GNU Radio algorithm with an embedded Python block enables on-the-fly waveform analysis and classification without requiring hardware-level reconfiguration.

The deep learning model combines CNN layers for feature extraction and LSTM networks for temporal sequence prediction, achieving a classification accuracy of 95.4% across 55 modulation schemes. Training data was generated using MATLAB’s Communication Toolbox for 1,10,000 waveform samples under various SNR and Eb/N₀ conditions.

The model was deployed using the ONNX framework to ensure cross-platform compatibility and integrated with SDR hardware (HackRF, USRP, Xilinx RFSoC PYNQ) for real-time testing. Evaluation metrics such as Bit Error Rate (BER) and throughput confirmed the system’s robustness under varying channel conditions.

Tech Stack Used: GNU Radio, HackRF/USRP/Xilinx RFSoC PYNQ, MATLAB, Python, TensorFlow, ONNX, CNN-LSTM Architecture, DVB-S2X Signal Dataset.

The Autonomous Navigation Aerial Vehicle (ANAV) was designed by Team XENITH from Kumaraguru College of Technology for the ISRO Robotics Challenge (URSC 2025). Built for autonomous operation in GPS-denied Martian conditions, the system integrates the NVIDIA Jetson Orin Nano and Pixhawk 2.4.8 flight controller to perform real-time sensor fusion, terrain mapping, and safe-spot landing.

Equipped with the MYNT EYE 3D Stereo Camera and SLAMTEC RPLiDAR S1, ANAV uses RTAB-Map SLAM, ICP, and Dijkstra’s Algorithm for precise localization and obstacle avoidance. A deep-learning module combining PointNet++ and ResNet detects safe landing zones, while an Emergency Response System ensures stability under power or link loss.

The drone achieved stable autonomous take-off, 30-second hover, and accurate landing during the qualification and elimination rounds, demonstrating advanced autonomy under Martian-like, GPS-denied conditions.

Tech Stack Used: Jetson Orin Nano, Pixhawk 2.4.8 (PX4), MYNT EYE 3D, RPLiDAR S1, ROS2, OpenCV, Python, RTAB-Map, Gazebo SITL, PointNet++, ResNet.

This project focuses on improving the efficiency, stability, and reliability of solar microgrids through data analytics, LSTM-based forecasting, and stochastic optimization techniques. The framework handles end-to-end processes from sensing to forecasting, optimization, simulation, and validation.

Appliance-level load patterns were extracted using a NILM algorithm trained with Harvard Dataverse datasets, enabling accurate demand-side modeling. Solar irradiance and load forecasts were developed using LSTM models.

The system optimizes tilt, panel orientation, inverter settings, and battery storage size using Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). PV–battery–load interactions were simulated in MATLAB/Simulink, while HOMER Pro and NREL SAM were used for sensitivity analysis, cost evaluation, and performance validation.

The framework also incorporates strategies to manage extreme conditions such as multi-day low sunlight, battery degradation, and system stress—using emergency SOC reserves, LFP chemistry, and robust energy management logic.

Tech Stack Used: MATLAB/Simulink, Python, LSTM, GA/PSO, NILM, HOMER Pro, NREL SAM, Harvard Dataverse.

Outcome: Achieved improved forecasting accuracy, optimized storage sizing, reduced energy losses, and a scalable intelligent framework suitable for rural, industrial, and smart microgrid deployments.

This project, carried out during my internship at Pricol Limited (Corporate Manufacturing Engineering Department), aimed to create a cost-effective vision-based color inspection system for wire harness verification. The system detects individual wire colors and validates their sequence to ensure compliance with product specifications during assembly.

The solution utilizes the Jetson Orin Nano Super Developer Kit paired with an Arducam IMX477 camera. Using OpenCV and HSV color space filtering, the system isolates wire colors in real-time, applies contour detection, and classifies each color accurately.

A PyQt5 GUI displays the live feed, detected colors, and sequence validation (“Correct” or “Incorrect”) in an intuitive interface. Detection logic is implemented through HSV thresholding and left-to-right positional sorting of identified colors.

The initial prototype was built using Raspberry Pi 5 with the Arducam IMX298 camera, later upgraded to Jetson Orin Nano for industrial-grade performance and GPU-accelerated processing. Future enhancements include multi-camera integration, AI-based classification for lighting variation handling, and PLC/SCADA connectivity for factory automation.

Tech Stack Used: Jetson Orin Nano, Arducam IMX477, OpenCV, Python, PyQt5, NumPy, HSV Color Model, GUI-based Vision System Design.

The Smart Glove-Based Gesture Control System was designed to replace traditional joystick or app-based control with a more natural, hands-free approach. The system uses an ESP32 microcontroller and an MPU6050 inertial measurement unit (IMU) to detect hand movements and transmit corresponding commands to the robot wirelessly over Wi-Fi.

The glove detects motion through accelerometer and gyroscope data from the MPU6050, processed by the ESP32 transmitter module. Commands such as forward, backward, left, right, and stop are sent to another ESP32 module on the robot, which drives the motors via an L298N motor driver.

The robot features Mecanum wheels that allow omnidirectional movement, offering precise motion control in all directions — forward, backward, diagonal, and rotational — without changing orientation. Real-time control is achieved through I²C communication between the ESP32 and IMU, ensuring low latency and smooth motion tracking.

The system’s flexible design makes it suitable for applications in assistive robotics, industrial automation, and search & rescue operations. Future enhancements include AI-based gesture prediction, computer vision integration, and autonomous path planning.

Tech Stack Used: ESP32, MPU6050, L298N Motor Driver, Mecanum Wheels, Python, Arduino IDE, I²C Protocol, Wi-Fi Communication.

The Pi Spy Rover is a Mecanum-wheeled robot built to demonstrate real-time omnidirectional mobility and remote surveillance. Unlike conventional robots, this design enables movement in any direction — forward, backward, sideways, diagonal, and rotational — without requiring reorientation, ideal for confined or dynamic environments.

The system integrates a Raspberry Pi with a Pi Camera for live video streaming over a Flask-based web interface. The ESP8266 microcontroller handles wireless communication and control commands, while an L298N dual H-bridge motor driver governs the speed and direction of four DC motors connected to Mecanum wheels.

A lightweight chassis supports all modules, powered by a 12V battery pack. The rover features WiFi-based control, allowing remote operation from any browser interface. It can perform complex motion such as lateral or diagonal translation and spot rotation (zero-radius turns), making it suitable for surveillance, research, and automation environments.

Future improvements include integrating computer vision for object detection, AI-based navigation, and obstacle avoidance for autonomous operation.

Tech Stack Used: Raspberry Pi, ESP8266, L298N Motor Driver, Flask (Web Interface), Python, OpenCV, Mecanum Wheels, Pi Camera, 12V Power Module.