Development of a real-time path recognition method for autonomous micro aerial vehicles
Standing
Undergraduate
Type of Proposal
Oral Presentation
Faculty
Faculty of Engineering
Faculty Sponsor
Dr. Mohammed Jalal Ahamed
Proposal
Micro Aerial Vehicles (MAVs) are increasing in popularity, specifically for indoor and outdoor navigation due to their compact size and high maneuverability. They can mitigate human danger in life threatening situations by autonomously navigating and traversing through dangerous environments, thus assisting in providing safe communities. Autonomous MAVs face two main challenges; GPS (Global Positioning System) restriction in indoor environments and a difficulty in accurate real-time path recognition for pin-point navigation. Previous research and development have been conducted to use an IMU (inertial measurement unit) combined with a real-time filtering algorithm for inertial sensor-based navigation in GPS restricted environments [1]. This research aims to develop a real-time detection algorithm for accurate path recognition. The algorithm proposed includes two mounted cameras on the MAV, one facing forward and the other facing downwards, to accurately construct a 3D model of the environment using machine vision techniques. The algorithm is then split into two separate scenarios; The MAV is in a corridor or the MAV is in a flight of stairs. When the MAV is in the corridor, the algorithm is specifically looking for vanishing points in the images obtained from the forward-facing camera. A vanishing point is a point on an image plane where two receding parallel lines in three-dimensional space appear to converge. Therefore, by searching for the highest populated vanishing point, we can then locate the end of the corridor to obtain the appropriate path for the MAV. When the MAV is in a flight of stairs, the algorithm is searching for sudden pixel differentiation to locate each step separately. By locating each step separately, we can then determine the total height and depth of the staircase and accurately detect an appropriate path for the MAV. As of right now, we are working on fully implementing this algorithm and incorporating it into the MAV. In the near future, we plan to combine this algorithm with the research done on the IMU’s real-time filtering algorithm to properly solve the main challenges an autonomous MAV faces. An autonomous MAV, when combined with filtered data from an IMU, and an accurate path recognition algorithm, offers an alternative for navigating and traversing through indoor environments. Such an autonomous MAV can aid in providing remote access to unknown or dangerous environments for assisting in gauging and proving safety.
References M. Shamoon, J. Jaekel, M. Ahamed, Development of a robust real-time filtering algorithm for inertial sensor based navigation systems with zero velocity update, UWill Discover, 2018.
Grand Challenges
Viable, Healthy and Safe Communities
Development of a real-time path recognition method for autonomous micro aerial vehicles
Micro Aerial Vehicles (MAVs) are increasing in popularity, specifically for indoor and outdoor navigation due to their compact size and high maneuverability. They can mitigate human danger in life threatening situations by autonomously navigating and traversing through dangerous environments, thus assisting in providing safe communities. Autonomous MAVs face two main challenges; GPS (Global Positioning System) restriction in indoor environments and a difficulty in accurate real-time path recognition for pin-point navigation. Previous research and development have been conducted to use an IMU (inertial measurement unit) combined with a real-time filtering algorithm for inertial sensor-based navigation in GPS restricted environments [1]. This research aims to develop a real-time detection algorithm for accurate path recognition. The algorithm proposed includes two mounted cameras on the MAV, one facing forward and the other facing downwards, to accurately construct a 3D model of the environment using machine vision techniques. The algorithm is then split into two separate scenarios; The MAV is in a corridor or the MAV is in a flight of stairs. When the MAV is in the corridor, the algorithm is specifically looking for vanishing points in the images obtained from the forward-facing camera. A vanishing point is a point on an image plane where two receding parallel lines in three-dimensional space appear to converge. Therefore, by searching for the highest populated vanishing point, we can then locate the end of the corridor to obtain the appropriate path for the MAV. When the MAV is in a flight of stairs, the algorithm is searching for sudden pixel differentiation to locate each step separately. By locating each step separately, we can then determine the total height and depth of the staircase and accurately detect an appropriate path for the MAV. As of right now, we are working on fully implementing this algorithm and incorporating it into the MAV. In the near future, we plan to combine this algorithm with the research done on the IMU’s real-time filtering algorithm to properly solve the main challenges an autonomous MAV faces. An autonomous MAV, when combined with filtered data from an IMU, and an accurate path recognition algorithm, offers an alternative for navigating and traversing through indoor environments. Such an autonomous MAV can aid in providing remote access to unknown or dangerous environments for assisting in gauging and proving safety.
References M. Shamoon, J. Jaekel, M. Ahamed, Development of a robust real-time filtering algorithm for inertial sensor based navigation systems with zero velocity update, UWill Discover, 2018.