ENGINEERING ARCHIVES
Autonomous Mobile Robots, Sim-to-Real Digital Twins, and Embedded Hardware Systems
Hybrid Vortex Crawler: Multi-Surface Wall-Climbing Robot
Multi-Surface Vertical Scaling & NDE Payload Delivery (NeX-Gen Robotics Challenge 2026 | IDREA)
Engineered an industrial wall-climbing inspection robot utilizing an active aerodynamic vortex 70mm Electric Ducted Fan (EDF) generating 45 N holding force combined with passive N52 magnetic track locking. Implemented a dual-tier control hierarchy: ROS 2 Jazzy on Raspberry Pi 4 for high-level mission logic and ToF telemetry, paired via UART to an ESP32 FreeRTOS controller for deterministic motor actuation and 50Hz tilt stabilization.
- Designed in Autodesk Fusion with 3mm CNC carbon fiber plates, Pololu 30T rubber tracks, and 4x 164 RPM planetary gear motors.
- Delivers 1.5 kg payload capacity for Non-Destructive Evaluation (NDE) ultrasound probes and HD thermal inspection.
- Dual-tier control: ROS 2 Jazzy mission supervisor (RPi4) coupled with FreeRTOS 20 kHz deterministic motor PWM (ESP32).
- Qualified for NeX-Gen Robotics Challenge 2026 (IDREA | Round 1 Concept Presentation).
AutoTwin-AI: Sim-to-Real Digital Twin
Zero-Defect Anomaly Detection in Automotive & Industrial Discrete Manufacturing
Proprietary Sim-to-Real Digital Twin pipeline eliminating the physical defect collection bottleneck. Ingests native 3D CAD files, generates 4,851 ray-traced domain-randomized synthetic renders under extreme illumination/camera noise, and trains an unsupervised PyTorch Autoencoder to localize micro-anomalies in sub-millimeter precision via reconstruction residual errors (L = ||X - X̂||²).
- Trained across 50 epochs on NVIDIA RTX 3050 Laptop GPU in ~128.8 min with MSE loss reaching 0.000092.
- Engineered FastAPI / TensorRT edge inference node delivering frame evaluation under 42ms.
- Constructed a 3-panel WebGL dashboard (Live Camera, AI Reconstruction, Residual Heatmap) for instant PLC line-trip alerts.
OOMWOO: Autonomous Robot Vacuum System
Authored core open-source coverage planning modules for the OOMWOO autonomous vacuum robot (oomwoo_clean_and_map). Implemented Boustrophedon Cellular Decomposition (BCD) for complete workspace coverage and integrated SLAM Toolbox for online map generation and localization under ROS 2 Jazzy.
Smart Traffic Management System (SIH25050)
Built a vision-based adaptive traffic signal controller utilizing YOLO object detection to dynamically calculate vehicular queue density across multi-lane intersections and modulate green-light intervals.
Hybrid AI Scholarship Policy Engine
Engineered an anomaly detection and policy verification bot designed to automate scholarship eligibility parsing, document fraud detection, and compliance auditing with an interactive Streamlit UI.
RESEARCH & PUBLICATIONS
Indexed Preprints, DOI Registrations & Multi-Agent RL Formulations
Sensor-Fusion Driven Deep Reinforcement Learning for Dynamic Traffic Signal Optimization
Author: Deepak R • Zenodo Preprint
This paper presents a closed-loop, decentralized urban traffic management architecture integrating real-time computer vision telemetry with multi-agent reinforcement learning (MARL). By ingesting optical density metrics and vehicle queue trajectories, the policy dynamically adjusts signal timing to maximize arterial throughput while minimizing waiting variance.
- Closed-loop telemetry pipeline fusing edge vision detections directly into policy state space.
- Multi-agent reward formulation preventing congestion spillover across adjacent intersections.
- Open access publication and dataset schema indexed under Zenodo DOI 10.5281/zenodo.20265628.
AURA: Acoustic-visual Urban Routing Architecture for Autonomous Mobile Robots
Author: Deepak R • Target: ICRA / IROS Conference Series
Investigating multimodal acoustic and visual sensor fusion frameworks to facilitate robust autonomous robot navigation and localized spatial awareness in GPS-denied, visually degraded urban canyons.
- Cross-modal acoustic-visual feature extraction for spatial landmark localization.
- Integration with ROS 2 Nav2 costmaps and real-time state estimators.
TECHNICAL LICENSES & CERTIFICATIONS
Industrial Robotics Career Programs, IBM Machine Learning & Linux Systems
Supervised Learning Methods
IBM
Data Analysis Using Python
IBM
Unsupervised Learning Methods
IBM
CTS Jan 26 Badge
Google Cloud Skills Boost
Digital Circuits
NPTEL
NPTEL25EE125S1154009277Industrial Robotics Theories for Implementation
NPTEL
NPTEL25ME161S1254009474Artificial Intelligence Fundamentals
IBM
Computer Hardware Basics
Cisco
Joy of Computing using Python
NPTEL
Weekly Coding Challenge 26
Unstop
3dc431aa-a8db-4827-9b74-40e00d356b9bCreate a Promotional Video using Canva
United Latino Students
MJL10F4I8EX0Certified Basic Resume Writer
TCS iON
Build a free website with WordPress
Coursera
RDD8RJ6MWZEUPearson MePro Level 10 Expert
Pearson
Introduction to Programming Using Python
upGrad
cb560194-63da-4524-898b-8d57b772a149