Researchers from the University of Hong Kong develop high-speed micro aerial vehicle
New research published in Science details how researchers at the University of Hong Kong have developed an advanced micro aerial vehicle (MAV) called SUPER, designed for high-speed navigation in unknown environments. Unlike previous attempts, which struggled with real-time obstacle avoidance at high speeds, SUPER integrates a lightweight lidar sensor and AI-driven control algorithms to navigate safely and efficiently.
The flying ability of birds has long inspired humans. Birds fly at high speeds through cluttered environments with minimal failures. Micro aerial vehicles (MAVs) like SUPER aim to replicate this agility for practical applications such as search and rescue. Achieving safe, high-speed flight in unknown environments requires a balance between agility, long-range sensing, and efficient trajectory planning using onboard processing.
Lidar – one of the keys to success
One of the keys to SUPER’s success is the Livox Mid-360 lidar sensor, a compact and lightweight device that provides precise, long-range obstacle detection. Manufactured by Livox, the Mid-360 offers a 360° horizontal field of view (FOV) and a 59° vertical FOV, enabling comprehensive environmental perception. It has a minimum detection range of 0.1 meters and can detect objects up to 40 meters away with 10% reflectivity. The sensor measures 65 mm × 65 mm × 60 mm and weighs 265 grams, making it suitable for integration into small aerial platforms. The current price of the Mid-360 is priced at $749.
Unlike traditional MAVs that primarily use vision sensors, which suffer from motion blur, short range, and low-light issues, SUPER leverages lidar for more reliable navigation. The system processes lidar point clouds in real time to detect open spaces, avoiding the computationally expensive occupancy grid maps used in other MAVs. Instead of relying on pre-calculated trajectories or external sensing, which limit real-world applications, SUPER uses an efficient planning framework that generates trajectories dynamically from lidar data, reducing mapping time to just one to five milliseconds.
Performance and application of SUPER
Tests showed that SUPER could fly at 20 meters per second through an obstacle course without incident. During trials in a dense forest, it successfully avoided trees, branches, and other obstacles while following a moving target, such as a human. Additionally, because its navigation system is based on lidar-inertial odometry, SUPER can operate independently in GPS-denied environments, making it ideal for real-world applications in exploration, logistics, and inspection. Traditional lidar-equipped aerial vehicles are often large, heavy, and slow due to bulky, costly sensors. SUPER overcomes these limitations with a compact, high-thrust-to-weight ratio design, using a lightweight Livox Mid-360 lidar for enhanced maneuverability.
To optimize trajectory planning, SUPER employs gradient-based methods instead of the slower mixed-integer quadratic programming, improving navigation speed and success rates. Its dual-trajectory system further enhances safety by maintaining a backup path while dynamically optimizing switching times. These advancements enable SUPER to surpass previous lidar-based UAVs in terms of speed, efficiency, and adaptability. Looking ahead, researchers suggest that smaller, lighter lidar sensors with longer ranges, improved aerodynamics, and enhanced motion prediction could further improve SUPER’s capabilities. With its low computational demands and high-speed safety features, SUPER is a promising solution for autonomous exploration, search-and-rescue missions, and operations in complex, unstructured environments.
For the full article in Science – Safety-assured high-speed navigation for MAVs
To read more about innovative drone design, Single Rotor Drone Spins for 360° LiDAR Scanning.
