A Depth-Based Hybrid Approach for Safe Flight Corridor Generation in Memoryless Planning.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Autonomous systems require reliable navigation in unknown environments.
• Local planning often relies on pre-existing maps, limiting real-time adaptability.
• Generating safe flight corridors is essential for collision avoidance.

Purpose of the Study:

• To develop a depth-based hybrid method for generating safe flight corridors.
• To enable memoryless local navigation planners to operate effectively.
• To ensure collision-free corridors with minimal overlap for path planning.

Main Methods:

• Utilizing raw depth images as direct input for a learning-based object-detection engine.
• Employing an object-detection network to predict polyhedral safe corridors.
• Implementing a verification procedure to guarantee collision-free corridors.
• Minimizing corridor overlap, achieving an average Intersection over Union (IoU) below 2%.

Main Results:

• Successful integration into a memoryless planner with a straight-line path-planning algorithm.
• High success rates demonstrated in both synthetic and real-world obstacle-dense environments.
• The method effectively produces separate, collision-free safe corridors.

Conclusions:

• The proposed depth-based hybrid method is highly capable for generating safe corridors for memoryless local planning.
• The approach eliminates the need for map fusion, enabling real-time navigation.
• This technique significantly enhances the safety and efficiency of autonomous flight in complex scenarios.