Vision based Autonomous Landing in Populated Environments a crucial aspect for urban deployment

Abstract

Safe autonomous landing in populated environments is a new and challenging open problem which will help to unleash the true potential of drones in applications in urban environments. Up to date, the use of drones has been limited to controlled scenarios, far awayfrom people, given the risk of accidents in case of system failure or human error, hence, more robust and resilient strategies, including autonomous emergency landing protocols, are indispensable in modern applications to guarantee the safe use of drones while operating in populated areas. Accordingly, we propose to use vision-based techniques along with modern Deep Learning approaches (Semantic Segmentation, Object Detection, Density Maps, etc.) to detect potential safe landing zones in unknown, unstructured and changingscenarios, like the ones encountered in urban areas. Semantic segmentation is proposed to classify different regions according to the level of risk of accidents, especially involving humans. Also, density maps will be used to locate people on the scene. Finally, a suitable landing zone should be selected trying to minimize the risk of accident, and the drone must land autonomously. For safetyreasons, the proposed strategy will be tested in realistic virtual environments with a real drone (robot-in-the-loop), before deployment in real scenarios.

Document Details

Document Type
DoD Grant Award
Publication Date
Jan 24, 2024
Source ID
N629092412001

Entities

People

  • Diego Mercado Ravell

Organizations

  • CIMAT Center for Mathematical Research
  • Office of Naval Research
  • United States Navy

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Aviation Safety Risk Assessment.
  • Computer Vision.

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Bayesian Inference
  • Autonomy