Aerial Robotic Workers: Design, Modeling, Control, Vision and Their Applications

Front Cover
Aerial Robotic Workers
Copyright
Contents
Contributors
1 Introduction
1.1 Introduction
1.2 Structure of the book
2 The fundamental hardware modules of an ARW
2.1 Introduction
2.2 Design of the ARWs
2.2.1 Frame design
2.2.2 Materials used for frame construction
2.2.3 Frame sizes and classification
2.3 Battery life
2.3.1 Different use cases
2.3.2 Battery characteristics
2.3.3 Battery selection process
2.4 Propulsion system
2.4.1 ARW flying and maneuvering principles
2.4.2 Propulsion system design
2.5 Sensor hardware
2.5.1 Vision sensors
2.5.2 Infrared ranging sensors
2.5.3 Inertial measurement unit
2.5.4 UWB ranging sensor
2.6 ROS node interfacing sensors and actuators
2.7 Compact manipulator for ARW
References
3 Modeling for ARWs
3.1 Notation
3.2 Introduction
3.3 Modeling of ARWs
3.3.1 Attitude modeling
3.3.1.1 Euler angles
3.3.1.2 Directional cosine matrix
3.3.1.3 Quaternion
3.3.2 Complete nonlinear model of a quadrotor ARW
3.3.3 Simplified position modeling
3.3.4 ARW foldable quadrotor modeling
3.4 Conclusions
References
4 Control of ARWs
4.1 Introduction
4.2 PID control
4.3 Linearized quadrotor model
4.4 LQR control
4.5 Linear model predictive control
4.6 Nonlinear MPC
4.6.1 Switching model based attitude controller of an ARW foldable quadrotor
4.7 Conclusions
References
5 Perception capabilities for ARWs
5.1 Introduction
5.2 Mapping
5.3 State estimation
5.3.1 Visual odometry
5.3.1.1 Feature extraction and matching
5.3.1.2 Pose estimation
5.4 Object detection and tracking
References
6 Navigation for ARWs
6.1 Introduction
6.2 Navigation architecture
6.3 Reactive local path planners
6.3.1 Artificial potential field
6.3.2 Nonlinear MPC with integrated obstacle avoidance
6.4 Global path planners
6.4.1 Map-based planning with D*+
6.4.2 Coverage path planner
6.4.2.1 Cooperative coverage path planner
6.4.2.2 Landmark-based path planner
6.4.3 Effective heading regulation
6.4.3.1 LiDAR-based heading regulation
6.4.3.2 The weighted arithmetic mean
6.4.3.3 Darkest contour heading regulation
6.4.3.4 Deepest point heading regulation
6.4.3.5 Centroid to heading rate command mapping
6.5 Conclusion
References
7 Exploration with ARWs
7.1 Introduction
7.2 The overall framework for exploration and path planning
7.2.1 Environment representation
7.2.2 Frontier based exploration
7.3 Combined path planning and exploration
7.3.1 The problem
7.3.2 ERRT solution
7.4 Conclusions
References
8 External force estimation for ARWs
8.1 Introduction
8.1.1 External force estimation
8.2 Disturbance rejection NMPC
8.3 Examples
8.3.1 MAV subject to wind-wall disturbances
8.3.2 Reducing the effect of a tethered payload
8.3.2.1 Example of center of gravity compensation
8.4 Conclusion
References
9 Perception driven guidance modules for aerial manipulation
9.1 Introduction
9.2 Perception architecture for ARW guidance
9.3 Object tracking
9.4 Object localization
9.5 Aerial manipulator system
9.5.1 Hexacopter manipulator carrier
9.5.2 The CARMA aerial manipulator
9.5.3 Visual sensor
9.6 Results
9.6.1 Object detection
9.6.2 Visual tracking
9.6.3 Stereo based guidance
References
10 Machine learning for ARWs
10.1 Introduction
10.2 AlexNet architecture
10.2.1 Transfer learning
10.3 Real-life scenarios
10.3.1 Junction detection
10.3.1.1 Data-set
10.3.1.2 Training and evaluations of the CNN
10.3.2 Human detection
10.3.2.1 FLIR ADAS thermal dataset
10.3.2.2 Training and the evaluation
10.3.3 Conclusions
References
11 Aerial infrastructures inspection
11.1 Introduction
11.2 Problem statement
11.3 Coverage path planning of 3D maps
11.3.1 Slicing and intersection
11.3.2 Clustering
11.3.2.1 Number of clusters
11.3.2.2 Graph clustering
11.3.2.3 Convex hull
11.3.3 Adding offset
11.3.4 Path assignment
11.3.5 Trajectory generation
11.3.6 Collision avoidance
11.4 Multiple agent visual inspection
11.4.1 Stereo mapping
11.4.2 Monocular mapping
11.5 Results
11.5.1 Simulation evaluation
11.5.2 Experimental evaluation
11.5.2.1 Experimental setup
11.5.2.2 Indoor artificial substructure inspection
11.5.2.3 Luleå University's fountain inspection
11.5.2.4 Wind turbine inspection
11.5.2.4.1 Wind turbine visual data acquisition
11.5.2.4.2 Tower coverage
11.5.2.4.3 Blade coverage
11.5.3 Quantification of the design choices
11.6 Lessons learned
11.6.1 ARW control
11.6.2 Planning
11.6.3 System setup
11.6.4 3D reconstruction
11.6.5 Localization
11.7 Conclusions
References
12 ARW deployment for subterranean environments
12.1 Introduction
12.2 Problem statement and open challenges
12.3 State estimation
12.4 Navigation and collision avoidance
12.4.1 Vision based methods
12.4.1.1 CNN classification
12.4.1.2 CNN regression
12.4.1.2.1 Centroid extraction
12.4.1.2.2 CNN architecture
12.5 Results
12.5.1 Experimental setup
12.5.2 Lidar based methods evaluation
12.5.2.1 Vector geometry based approach
12.5.2.2 Weighted arithmetic mean approach
12.5.3 Vision based methods evaluation
12.5.3.1 Darkness contours detection
12.5.3.2 CNN classification approach
12.5.3.3 CNN regression approach
12.6 Lessons learned
12.6.1 ARW control
12.6.2 Localization
12.6.3 Navigation
12.7 Conclusion
References
13 Edge connected ARWs
13.1 Introduction
13.2 Edge computing
13.2.1 Cloud – Fog – Edge
13.2.2 Virtual machines – containers – kubernetes
13.2.2.1 Virtual machines
13.2.2.2 Containers
13.2.2.3 Kubernetes
13.3 Communication layer
13.4 Conclusion
References
Index
Back Cover