Towards Autonomous Miniature Rotorcrafts in Cluttered Environments for Scene Understanding
Abstract
The objective of this project is to develop control, navigation, computer vision, and 3D mapping algorithms for Micro Aerial Vehicles (MAVs) to autonomously explore obstacle-dense environments by fusing multiple sensors (camera, Omni-stereo vision system, RGB-D sensor, IMU, laser scanner, etc.) for 3D scene understanding. We envision that the autonomous MAVs can be used in various military scenarios, such as surveillance and reconnaissance in urban or wooded environments to enhance tactical situational awareness or in cave search applications to build a 3D map. During the course of the project, we have made significant accomplishments in tackling the critical challenges of MAV perception, control and 3D SLAM in unknown environments and verify the proposed methods on the experimental test-bed (i.e., quadrotor MAV from AscTec). The achievements include the development of single camera omni-stereo vision system, real-time pose estimation via multimodal sensing (i.e., laser scanner, RGB-D sensor, and Google Tango tablet, etc.), fast visual odometry and 3D mapping algorithms, 3D path planning algorithms, quaternion-based orientation estimation using low-cost Inertia/Magnetic sensors, etc. The success of this project has taken us a few steps closer to empower the MAVs with the similar exploration capability of flying birds to achieve autonomous navigation in cluttered environments. This final report summarizes the significant technical achievements of the project.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 17, 2018
- Accession Number
- AD1069020
Entities
People
- Jizhong Xiao
- Zisis I. Petrou
Organizations
- City College of New York