Monocular Visual-Inertial SLAM-Based Collision Avoidance Strategy for Fail-Safe UAV Using Fuzzy Logic Controllers

Abstract

In this paper, we developed a novel Cross-Entropy Optimization (CEO)-based Fuzzy Logic Controller (FLC) for Fail-Safe UAV to expand its collision avoidance capabilities in the GPS-denied environments using Monocular Visual-Inertial SLAM-based strategy. The function of this FLC aims to control the heading of Fail-Safe UAV to avoid the obstacle, e.g. wall, bridge, tree line et al, using its real-time and accurate localization information. In the Matlab Simulink-based training framework, the Scaling Factor (SF) is adjusted according to the collision avoidance task firstly, and then the Membership Function (MF) is tuned based on the optimized Scaling Factor to further improve the control performances. After obtained the optimal SF and MF, 64 % of rules has been reduced (from 125 rules to 45 rules), and a large number of real see-and-avoid tests with a quadcopter have done. The simulation and experiment results show that this new proposed FLC can precisely navigates the Fail-Safe UAV to avoid the obstacle, obtaining better performances compared to only SF optimization-based FLC. To our best knowledge, this is the first work to present the optimized FLC using Cross-Entropy method in both SF and MF optimization, and apply it in the UAV.

A comparative study on the control of quadcopter UAVs by using singleton and non-singleton fuzzy logic controllers

“…Although there are a number of FLC applications for navigating a quadcopter UAV in literature, e.g., [11], most of these are SFLCs, which focus on high level navigation, rather than exploring the effect of various levels of input uncertainty on the control performance. On the other hand, it is reported in literature that NSFLCs give more promising results when compared to their singleton counterparts for non-linear servo system [12], chaotic time series prediction [13], etc.…”

Geo-Mapping and Visual Stitching to Support Landmine Detection Using a Low-Cost UAV

“…The total mass of the robot is 900Kg with a length of 4m. Other platforms such as the Titan-IX [9] and the SILO6 [10] show that legged-robots are able to navigate more precisely over roughly terrains than wheeled-robots. Nonetheless, both wheeled and legged mechanisms are simply too expensive and heavy to be used within the Colombian context, where the military must transport tons of equipment to difficult access locations.…”

“…The bottom camera has a field of view of 47.5 o with a resolution of 320 × 240 pixels at 60 fps. Despite the hardware and onboard processing limitations of this robot, the AR.Drone 2.0 has been used in several research projects in control [10], autonomous navigation [7], image processing [14] and swarm robotics, due to its low cost and robust mainframe. In this project, we installed a commercial GPS sensor onboard the AR.Drone quadrotor called Flight Recorder.…”

“…3: Defining the a located aroun 4: Analyzing the 5: if there are at I n > I p + t or 6: The point p 7: else 8: The point p I n is the bri the selected 9: end if 10 where R and T are the rotation and translation matrix of the UAV, N is a normal vector to the ground plane, d denotes the distance between the UAV and the ground plane, and K denotes the intrinsic parameters of the camera. These parameters are described as follows:…”

Similarity-based non-singleton fuzzy logic control for improved performance in UAVs

“…They are able to deliver adequate control and handle uncertainties without the requirement of an accurate mathematical UAV model. Among the different types of FLCs, singleton FLCs (SFLCs) are the most common FLCs used for the UAVs [11]. However, SFLCs are not capable of capturing input uncertainties effectively.…”