NEXNAV - NexisEdge Navigation System Short, brandable, and directly connected to your team. Ideal for a long-term UAV navigation platform.
GPS-independent autonomous drone navigation using triangulation, A\* pathfinding, and intelligent path planning for obstacle-aware flight.
- Problem, approach and result explained
- Public source code
- Expert reviewed· 6 reviews
82%
Core navigation algorithms implemented A*, triangulation-based positioning and autonomous path planning.
93%
Environmental scanning capability RPLIDAR A1 incorporated into the sensor architecture.
89%
Directional distance sensors Two TFmini Plus LiDAR sensors in the hardware configuration.
Overview
The problem
Conventional autonomous drones heavily depend on GPS for positioning, navigation, and mission execution. However, in environments such as indoor facilities, tunnels, dense forests, disaster-hit zones, underground structures, and areas affected by GPS jamming or signal interference, GPS signals can become unavailable or unreliable. This creates a major challenge for autonomous UAV operations, particularly in surveillance, search and rescue, disaster management, and critical infrastructure inspection, where accurate positioning and reliable navigation are essential.
The problem directly affects emergency response teams, defence and surveillance agencies, industrial inspection operators, and drone developers who require reliable autonomous flight in GPS-denied environments. Without GPS, drones may lose their positional reference, struggle to determine their location, and become incapable of navigating safely around unknown obstacles.
Our objective is to develop an intelligent GPS-independent autonomous navigation system that enables drones to estimate their position, understand their surroundings, and calculate optimal navigation paths without depending on conventional GPS signals. We aim to achieve this using LiDAR-based environmental sensing, triangulation-based positioning, and the A* pathfinding algorithm, integrated with an onboard computing platform and flight controller. The ultimate goal is to enable reliable autonomous navigation, real-time obstacle detection, dynamic path replanning, and safe flight execution in GPS-denied environments.
The approach
We developed a GPS-independent autonomous drone navigation framework using a Raspberry Pi 4 (4GB) as the companion computer and a Pixhawk flight controller running ArduPilot for flight management. Our sensing architecture incorporates an RPLIDAR A1 360-degree LiDAR and two TFmini Plus distance sensors to support environmental perception and obstacle-distance measurement.
On the software side, we have implemented three core navigation components. First, the A* (A-Star) pathfinding algorithm calculates an optimized route by evaluating the actual travel cost and estimated remaining distance to the destination using f(n) = g(n) + h(n). Second, we implemented a triangulation-based positioning method to support GPS-independent position estimation using geometric measurements. Third, we implemented the autonomous path-planning module to generate navigation routes based on the available positional and environmental information. We are also developing DristiGCS, our ground control station, to provide centralized telemetry monitoring, flight controls, and navigation visualization.
One of our hardest engineering decisions was to avoid treating GPS replacement as simply installing another positioning sensor. Instead, we chose a sensor-driven navigation architecture that combines geometric positioning, environmental perception, and algorithmic path planning. We also decided to prioritize individual sensor validation and algorithm verification before attempting complete autonomous flight integration. This reduces integration risks and makes it easier to isolate hardware and software failures.
We deliberately did not claim full autonomous obstacle avoidance or closed-loop flight execution at this stage. Although A*, triangulation-based positioning, and autonomous path planning have been implemented, automatic flight actions when obstacles are detected, dynamic path replanning using live sensor feedback, and execution of navigation commands through Pixhawk are still under development. Our next step is to connect the sensor-processing pipeline with the navigation algorithms and flight controller, validate the complete feedback loop in simulation, and then proceed to controlled hardware testing.
The result
The project has successfully progressed from a conceptual GPS-denied navigation architecture to an implemented software prototype with three major navigation capabilities: A* pathfinding, triangulation-based positioning, and autonomous path planning. The A* algorithm can calculate optimized routes between defined starting and destination points using a cost-based search approach. The triangulation-based positioning method has been implemented to support position estimation without conventional GPS dependency. Additionally, the autonomous path-planning module has been developed to generate navigation routes based on the available positional and environmental information.
On the hardware side, we have established the Raspberry Pi 4 as the onboard computing platform, with a Pixhawk running ArduPilot as the flight controller. Our sensor architecture incorporates one RPLIDAR A1 360-degree LiDAR and two TFmini Plus distance sensors. Sensor-level testing and hardware integration are currently underway. We have also developed the initial DristiGCS ground control software to support vehicle monitoring and flight-system interaction.
The most significant result so far is the successful implementation of the core computational foundation required for GPS-independent navigation. However, real-time obstacle-responsive flight behaviour, dynamic path replanning, and autonomous execution of generated routes through the Pixhawk have not yet been completed. Therefore, full autonomous flight performance has not yet been established.
We are evaluating the system against a conventional GPS-dependent navigation baseline and a non-adaptive predefined-route approach. The planned evaluation metrics include positioning error, route length, path-planning computation time, obstacle-detection response latency, successful obstacle avoidance rate, and autonomous mission completion rate. These measurements will be collected through controlled simulation and hardware experiments. Our current achievement is the implementation of the core navigation algorithms, with the next milestone being their integration into a complete real-time autonomous navigation and obstacle-avoidance system.
Reflection
This project taught us that implementing navigation algorithms is only one part of achieving true autonomy. While A*, triangulation-based positioning, and autonomous path planning have been implemented, real-time obstacle avoidance, dynamic path replanning, and autonomous flight execution remain unfinished. We learned the importance of sensor accuracy, data synchronization, and seamless integration between perception, planning, and flight control. If we started again, we would prioritize sensor validation and simulation-based testing earlier. Our next focus is completing hardware integration, implementing real-time obstacle response, and validating autonomous navigation through SITL and controlled flight tests.
Submitted to 3D Autonomous Path Planning
- Your idea in one line
A GPS-denied autonomous UAV that uses 360° LiDAR, TFmini Plus sensors, and optical flow for real-time obstacle avoidance and intelligent navigation without GPS.
- What will be working by the finale?
By the finale, a working autonomous drone will demonstrate GPS-denied navigation, real-time obstacle detection using LiDAR and TFmini Plus sensors, and autonomous obstacle avoidance using onboard processing and optical flow for stabilization.
- Tech stack and hardware you plan to use
Tech Stack: Python, C++, ArduPilot, MAVLink/PyMAVLink, ROS 2 Jazzy, SLAM, A* path planning, FastAPI, React, Vite, Material UI, and ArduPilot SITL. Hardware: Raspberry Pi 4 (4GB), Pixhawk flight controller, Slamtec RPLIDAR A1 (360°), 2 × TFmini Plus LiDAR sensors, optical flow sensor, brushless motors, ESCs, and LiPo battery.
- What have you built or tested so far?
Tested all the sensors mounted over the drone , Sensor Fusion has been done, tested collection of data to rpi and fetching to the dashboard
- What's working now?
The GPS-Denied Drone project has made significant progress in the development of its autonomous navigation and path-planning architecture. The *A (A-Star) algorithm has been successfully implemented*\* for intelligent pathfinding and optimal route generation. Additionally, the **triangulation-based positioning method has been implemented** to support position estimation without relying on GPS. The **autonomous path-planning module has also been implemented**, enabling the system to calculate navigation paths based on the available environmental information. However, the actual execution and response mechanisms are still under development. Specifically, automated actions such as dynamically changing the flight path when an obstacle is detected, avoiding obstacles in real time, and executing autonomous flight commands based on navigation decisions have not yet been implemented. The current focus is on integrating the implemented algorithms with real-time sensor data and the Pixhawk flight controller to establish a complete autonomous navigation and obstacle-response system.
- GitHub repository
- github.com/nexis-edge/Bput_NE_003.git
- Deck or report (PDF)
- Deck or report (PDF)PDF
AI tools used
ChatGPT for Preparing Report, and Testing