Research Works and Projects
Research Works and Projects
Embedded Files
DynoLoc: Zero-Infrastructure 3D Tracking for Emergency Response
DynoLoc: Zero-Infrastructure 3D Tracking for Emergency Response
Reliable indoor localization is a critical bottleneck in emergency response like in the case of firefighters. To address this, I designed a rapidly deploy-able 3D tracking solution that requires zero pre-existing infrastructure. By fusing Ultra Wide Band (UWB) radio, Long Range (LoRa) radio and Inertial Measurement Unit (IMU) data into wearable 'tags,' the system—DynoLoc—allows first responders to establish a localized tracking grid instantly upon arrival in a multi-storied building with non-line-of-sight. The system allows incident commanders to remotely monitor personnel movements in real-time, facilitating safer navigation through hazardous and un-mapped spaces. The project demonstrates how low-power wide-area networking and precision radio ranging can be synthesized to improve operational safety and mission efficiency.
The DynoLoc prototype leverages Commercial Off-The-Shelf (COTS) components to ensure scalability and cost-effectiveness. By integrating the Qorvo (formerly Decawave) UWB module, the system achieves high-precision sub-meter 3D localization across expansive indoor environments. To overcome line-of-sight limitations and further enhance spatial coverage, the system supports an ad-hoc deployment model; users can strategically place inexpensive secondary tags to extend the tracking perimeter and achieve sub-meter accuracy in complex signal environments
The Tech: A multi-modal fusion of UWB (precision), IMU (motion), and LoRa (long-range communication), pressure sensor and other secondary sensors like hazardous gas detector, camera etc.
The Deployment: 'Drop-and-go' architecture; a single anchor establishes a 3D coordinate system at the scene entry.
The Goal: Real-time remote monitoring of personnel in hazardous, unexplored indoor environments.
The Impact: Enhanced situational awareness for commanders, leading to faster rescue operations ( few seconds vs tens of minutes to locate in fire drill in realistic settings) and increased safety for first responders
Published Paper: [ Dynoloc PDF ] Code: [ github ]
FSO-VR: Steerable Free Space Optics Link for Virtual Reality Headset
FSO-VR: Steerable Free Space Optics Link for Virtual Reality Headset
FSO-VR is a steerable Free Space Optics (FSO) system designed to provide a high-bandwidth, wireless line-of-sight link between a fixed transceiver and a mobile VR/AR headset. To maintain a constant connection during user movement, the system utilizes a Galvo-Mirror (GM) collimator that dynamically steers the beam based on real-time RSSI feedback transmitted via a secondary radio channel.
Key contributions include:
Data-Driven Steering: Developed a novel beam-steering algorithm optimized through the analysis of user kinematics in various VR gaming environments.
Rigorous Simulation: Conducted extensive link-budget analysis using Zemax to evaluate performance across diverse angular displacements.
Hardware Prototyping: Validated the architecture using COTS components, including ThorLabs precision galvos, TimberCon optical sensors, and Oculus hardware.
The result is a truly "cord-cut" experience, enabling seamless, high-mobility wireless connectivity capable of supporting demanding 8K VR stream
Published Paper: [ FSO-VR PDF ]
FSONet: Reconfigurable Backhaul for Urban Cellular Network
FSONet: Reconfigurable Backhaul for Urban Cellular Network
FSONet is an end-to-end cellular backhaul architecture that leverages steerable Free Space Optics (FSO) to enable dynamic network reconfiguration. My work spans the full design cycle: I developed a robust 100m hardware prototype featuring a novel Tracking and Pointing (TP) mechanism driven by photodiode signal feedback. To ensure reliability, I conducted extensive link budget and TCP analyses, proving the system’s resilience against platform vibration, atmospheric turbulence (fog), and temporary occlusions—all while maintaining eye-safety compliance. On the networking layer, I devised a near-optimal transceiver placement strategy for 3D urban topologies (e.g., NYC, San Francisco), enabling uninterrupted, on-demand flow routing in complex environments.
Hardware: Developed a 100m FSO link prototype with a custom Tracking and Pointing (TP) system utilizing receiver-end photodiode feedback for sub-degree alignment.
Resilience: Validated link robustness against environmental stressors (fog, vibration) and temporary signal blockage via comprehensive link budget and TCP recovery analysis.
Optimization: Formulated a near-optimal transceiver placement and flow-routing algorithm, evaluated using 3D building data from major metropolitan areas to ensure high-capacity, uninterrupted backhaul performance.
SpecSense: Crowdsensing for Dynamic Wireless Spectrum Access
SpecSense: Crowdsensing for Dynamic Wireless Spectrum Access
Radio spectrum is a finite and increasingly congested resource. This project introduces a comprehensive crowdsensing architecture designed to enable Dynamic Spectrum Access (DSA) by accurately mapping occupancy in real-time. My research followed a three-tiered approach: first, I developed an efficient querying and interpolation framework to generate high-fidelity occupancy maps using low-cost sensors (RTL-SDRs). Second, I implemented a temporal clustering technique to track time-varying usage patterns, validated through terrain-aware propagation simulations. Finally, I proposed an opportunistic sharing model that accounts for primary receiver locations to prevent interference. Together, these contributions facilitate seamless, high-efficiency spectrum sharing between primary and secondary licensees.
Accurate Mapping: Developed spatial interpolation techniques using low-cost sensors (RTL-SDR) to visualize spectrum voids.
Temporal Analysis: Created clustering algorithms to generate time-varying occupancy maps based on real-world terrain propagation.
Interference Mitigation: Designed an opportunistic sharing model that protects primary transmitters by estimating receiver locations.
Result: A unified architecture that maximizes spectrum utility and enables on-demand access across diverse frequency bands.
Virtual Network Embedding in Reconfigurable Networks
Virtual Network Embedding in Reconfigurable Networks
This project provides the first formal treatment of Virtual Network Embedding (VNE) in reconfigurable networks. Two demand models for data traffic are presented - traditional fixed-bandwidth and stochastic-bandwidth model designed to maximize utilization. A runtime-binding embedding reconfigures the links dynamically according to network state. Using real-world datacenter traffic statistics, we show that our reconfigurable models offer a 30–40% performance gain over existing baselines.
Novel Framework: The first VNE solution tailored for reconfigurable network topologies.
Stochastic Modeling: A new demand model that leverages runtime reconfiguration for better utilization.
Real-world Validation: Demonstrated 30-40% gains using actual datacenter traffic traces.
Page updated
Google Sites
Report abuse