Technical Experiments Summary

This repository contains six key experiments conducted to evaluate and improve various aspects of the Intelligent Flight Tracking Assistant system. Below is a summary of each experiment and its key findings.

---

✅ Experiment 1: Improving Rendering Performance When Displaying 10,000+ Objects

• Goal: Evaluate performance degradation when rendering over 10,000 objects and its impact on UI responsiveness.

• Approach: Implemented and benchmarked multiple OpenGL rendering methods (e.g., VBO) using C#.

• Result: VBO showed the best rendering time (0.062s) among tested methods.

• Conclusion: VBO was selected as the optimal rendering approach under OpenGL 2.x constraints.

📄 View Details

---

✅ Experiment 2: Network Connection Recovery Experiment

• Goal: Verify whether disconnection and reconnection can be reliably detected and recovered using ping/echo and TCP methods.

• Approach: Tested ICMP and TCP-based monitoring with scheduled connection checks and measured recovery timing.

• Result: ICMP was blocked; TCP-based monitoring succeeded with 100% success in 50 trials for both internet and USB disconnection.

• Conclusion: TCP-based periodic connection attempts are reliable for fault detection and enable timely recovery within 1 minute.

📄 View Details

---

✅ Experiment 3: Implement a C#-based prototype

• Goal: Validate whether a C#/WPF implementation can match or exceed the functionality and performance of the existing system.

• Approach: Reimplemented core features including UI, TCP communication, OpenGL rendering, and map display using C# and WPF.

• Result: All key components were successfully implemented with performance comparable to the current C++ version.

• Conclusion: The C#/WPF-based prototype proved feasible and was confirmed as a viable replacement for the legacy implementation.

📄 View Details

---

✅ Experiment 4: Performance Test for Aircraft CPA (Closest Point of Approach) Computation

• Goal: Evaluate the computational load of CPA (Closest Point of Approach) calculations for large-scale aircraft tracking data.

• Approach: Conducted pairwise CPA calculations using ~9,000 recorded aircraft and compared full-set vs. distance-filtered computation.

• Result: Full CPA computation took ~13.5s on average; applying an 85NM horizontal filter reduced time to ~0.59s.

• Conclusion: Pre-filtering based on distance is essential for reducing computation time and enabling real-time CPA analysis.

📄 View Details

---

✅ Experiment 5: SDR connection recovery experiment

• Goal: Verify automatic detection and recovery from physical disconnection or malfunction of SDR or antenna on a Raspberry Pi system.

• Approach: Developed systemd services and a custom sdrmonitor script to manage dump1090 and ADSBHub feeding based on hardware state.

• Result: All 50 manual disconnect/reconnect tests and 49/50 reboots succeeded in automatic recovery; 1 reboot failed due to unrelated OS issue.

• Conclusion: The proposed tactic demonstrated reliable fault detection and full recovery, ensuring robust ADS-B data feed availability.

📄 View Details

---

✅ Experiment 6: Polygon-based Aircraft Filtering Module Implementation

• Goal: Determine if aircraft positions (lat/lon) can be efficiently filtered against polygonal areas in real-time.

• Approach: Implemented and benchmarked filtering logic in Python, C++, and C# using both ENU and lat/lon comparisons.

• Result: C++ filtering was ~14× faster than C# (0.5ms vs. 7ms); lat/lon-based filtering was successfully integrated and validated in the project.

• Conclusion: C# is suitable for integration, with a C++ DLL recommended for performance-critical scenarios.

📄 View Details

---

📌 For architecture decisions related to these experiments, please refer to the associated ADRs and design documents in the architecture repository.