Diganta
no gps and no external beacons only the sensor on the drone
- Video of it working
- Expert reviewed· 6 reviews
Walkthrough
Overview
The problem
3D Autonomous Drone Path Planning in GPS-Denied Areas refers to developing a system that enables a drone to navigate autonomously in three-dimensional space (length, width, and height) without relying on GPS signals.
In normal drone operations, GPS provides information about the drone's geographical position. However, in GPS-denied environments, such as indoor spaces, tunnels, dense forests, or areas where GPS signals are unavailable, the drone must rely on onboard sensors to estimate its position, understand its surroundings, and determine its flight path
The approach
We plan to solve this problem by developing a GPS-independent navigation system using onboard sensors such as IMU and optical flow. First, we will calibrate these sensors to obtain accurate motion measurements. Then, we will combine their data using sensor fusion techniques to estimate the drone's position and movement without GPS. Using this information, along with obstacle detection, we aim to implement 3D path planning so that the drone can identify obstacles, calculate safe routes, and navigate autonomously. Our existing flight algorithm has already been tested in simulations, and our current focus is on achieving reliable real-world implementation
Submitted to 3D Autonomous Path Planning
- Your idea in one line
The drone autonomously follows a user-defined route using UWB anchors, optical flow and an Extended Kalman Filter for improved accuracy, a barometer and ToF sensor for altitude estimation, and AprilTags for precise docking, returning to the base station after completing its mission.
- What will be working by the finale?
The autonomous docking with charging station and UWB anchor system.
- Tech stack and hardware you plan to use
Pixhawk 2.4.8 flight controller, F450 quadcopter frame, Radxa Zero 2 Pro companion computer, UWB modules, optical flow sensor, camera module, 2D LiDAR (optional), barometer, ToF sensor, telemetry and RC receiver; Using Python and C++ for autonomous navigation, sensor fusion, and control, MAVLink for communication, QGroundControl for monitoring, and Git/GitHub for version control.
- What have you built or tested so far?
We have developed a drone simulator that integrates drone physics, environmental configuration, MAVLink communication, flight recording, and control mechanisms. We implemented a Mission Planner to define autonomous flight paths using X, Y, and Z coordinates, set altitudes, and configure waypoint commands such as Hold, Rotate, Land, and Return to Launch (RTL). We also incorporated obstacle configuration, UWB-based positioning, and live 3D odometry and SLAM visualization. The autonomous flight demonstration shows the drone following its user-defined path, detecting and avoiding obstacles, returning to the planned trajectory, executing waypoint commands, maintaining positioning accuracy, and reaching its landing point. We also developed a flight data analysis workflow using a Python script to convert flight logs and path files into graphs for evaluating positioning errors, precision, speed, climb rate, and the actual trajectory followed by the drone. These analyses help us examine flight performance and compare the actual path with the predefined route. Beyond simulation, we demonstrated the physical drone's ability to take off, reach and maintain a desired altitude, fly stably, descend under throttle control, identify a suitable landing position, and land safely. Overall, we achieved an integrated workflow for autonomous drone simulation, mission planning, obstacle avoidance, flight performance analysis, and real-world validation of key flight-control capabilities.
- Pitch deck (PDF)
- Pitch deck (PDF)PDF
- Demo video
- youtu.be/eZfNsuRLDfI
- What's working now?
Our current focus is on transitioning the drone from manual control to fully autonomous flight. At this stage, we are primarily working on calibrating the autonomous drone system to ensure accurate and stable operation. One of the major challenges we are encountering is the calibration of the optical flow sensor. Since our system relies heavily on optical flow for movement and position estimation, the absence of GPS makes the calibration process more difficult. Although we have successfully developed the autonomous flight algorithm and obtained satisfactory results through simulations, implementing it on the actual drone remains a challenge. Our primary objective now is to establish a reliable calibration process that enables the drone to achieve stable and autonomous flight without depending on GPS.
AI tools used
for fast development