A contract based approach to collision avoidance for UAVs
Loading...
Authors
Alimbayev, Talgat
Issue Date
2020-08
Type
Electronic thesis
Thesis
Thesis
Language
ENG
Keywords
Mechanical engineering
Alternative Title
Abstract
Unmanned Aerial Vehicles (UAV) are becoming widespread and making them autonomous is the next big milestone. To control UAVs, a Guidance-Navigation-Controller (GNC) architecture is commonly used. The UAV subsystems may experience changes in their performance levels. These changes can negatively affect the performance of the UAV in some critical scenarios such as the obstacle avoidance. First, modifications to the conventional GNC architecture were proposed in order to have the information about the current level of performance of the UAV subsystems. At this point, ways to measure the performance of the Navigation and Controller subsystems were suggested. After that, a methodology to incorporate this information about the changing performance levels of the Navigation and Controller subsystems into the UAV path planning was proposed. To do so, a contract based reasoning is employed and the UAV is decomposed into subsystems. Under the contract based reasoning each subsystem guarantees a certain level of its performance and assumes a certain level of performance from other subsystems as well. The contract between the subsystems is then formed which is an additional inequality constraints added to the path planning algorithm in the Guidance subsystem. The contract is needed to assure the safe UAV operation under changing performance levels of the subsystems. The Guidance subsystem is the one that enforces satisfaction of the contract. Then, the proposed methodology was demonstrated in the obstacle avoidance scenario for a simulation. Finally, it was shown through a simulation how the proposed methodology could possibly work for an airborne UAV through an artificial set of subsystem performances and operational environments.
Description
August 2020
School of Engineering
School of Engineering
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY