90319 Update:
 UAS (Unmanned Aerial Systems)
 UAV (Unmanned Aerial Vehicle)
 Introduction to Autonomous Flight
 History
 Morphologies
 Components
 Airframe
 ESC (electronic speed controllers) motor speed control
 Propellers
 Batteries
 Attitude Control
 Balance of Net Forces – Hovering
 Unequal Net Force – Directional Movement
 Autopilot
 Sensors – MEMS (micro electronic mechanical systems)
 IMU Gyros (measure relative change in attitude) 3ea.
 IMU accelerometers 3ea.
 GPS
 Other sensor devices
 Cameras
 Lazer Range Finders
 Compasses
 Barometers
 Flight Computer – responsible for higher level functions
 High Level Functions at roughly 5 hz. per second
 Position Control Loop
 Flight Computer
 Using GPS
 Calculates the Required Target Thrust Vector
 Sends inputs to the Attitude Control Loop
 Position Control Loop
 Lower level functions at roughly 50 hz. per second
 Attitude Control Loop
 Autopilot
 IMU
 Motors
 Attitude Control Loop
 High Level Functions at roughly 5 hz. per second
 Search Planning
 Planning Factors / 2 D Search Space
 Includes Dynamic Vehicular Contraints
 Search Space Discretization
 Possible paths
 Start location
 Goal location
 Actions (L,R,U,D,Angular)
 Costs
 Cost functions enable plan comparisons
 Path representations
 Grids Size Tradeoffs
 Partial Plans as Part of the Planning Process
 Generate a List of Visited Actions and Obstacles
 Types of Searches
 Breadth First Search
 Always Expands Shortest Current Plan First
 Guaranteed to Find Shortest Path First
 Can be Computationally Expensive Because it must keep a number of partial plans in memory / storage while it is expanding all new plans
 DepthFirst Search
 Expands Last Successful Action
 Requires a Bit of Luck
 You never know if this is possibly the shortest plan
 Can Fail Completely
 Neither of the above plans make use of a map
 Uniform Cost Search
 Expands the Plan with the Lowest Cost First
 Guaranteed to Find the Lowest Cost Plan First
 Types of Costs
 Euclidean Distance = sqrt(sum(x^2 + y^2))
 Manhatten Distance = sum of x & y distances remaining to get to the goal
 Heuristics Admissible and Consistent
 Admissibility = always needs to be an underestimate of the true cost
 Ignoring the cost to go around obstacles ensures admissibility
 G = cost function (i.e., sum of the actions taken so far)
 H = heuristics (i.e., underestimate of the remaining cost to get from the last state of a partial plan to the goal)
 F = G + H (estimate of total cost)
 Consistent
 Triangle inequality theorem where: H (x1 > x3) <= H(x1 > x2) + H(x2 > x3)
 A star or A*
 expanding the partial plan that has the lowest total cost in terms of the sum of the actual costs of actions in the partial plan plus the heuristic from the last state in the partial plan is the A* algorithm
 Guaranteed to find the lowest cost plan first
 A* search doesn’t suffer the inefficiencies of Breadth First Search in terms of memory or storage of large numbers of partial plans during execution but is dependent on the quality of the chosen heuristic.
 Breadth First Search
 Grids to Graphs
 Graph methods reduce memory overhead
 Waypoint Extraction
 Collinearity
 p1 = (x1, y1, z1)
 p2 = (x2, y2, z2)
 p3 = (x3, y3, z3)
 Area = det(matrix([p1,p2,p3])) = 0 = colinear
 Area = x1(y2 – y3) + x2(y3 – y1) + x3(y1 y2)
 Discretization
 Search from start to goal
 Test for collinearity & remove elements or cells
 Ray Tracing
 Bresenham
 Grids to Graphs
 Graph Tradeoffs
 Medial Axis
 Voronoi Graphs
 Graph Search
 Dead bands
 Collinearity

 3D Grids
 Voxel maps
 2D Environment
 2D maps
 Pixels = picture elements 2D
 Planning Factors 3D Search Space
 3D Environment
 2.5D maps
 Random Sampling
 Probabilistic Roadmap
 Local Planning
 Receding Horizon
 Replanning
 Real World Planning
 Constraints
 Modeling Dynamics
 Dubins Car
 Steering
 RRT
 Adding Obstacles
 Potential Field Planning
 Voxels = 3D elements
 Voxelmap
 kdtree
 knearest neighbors
 Probabilistic road map
 Receding horizon path planning



 Planning as search
 Astar search algorithm
 Dubinscar
 Steering
 RRT (rapidly exploring random tree)
 Potential field
 Path pruning
 Collinearity
 Bresenham
 Medial Axis
 Voronoi diagrams
 Graph search
 BFS
 DFS
 Planning as search




 Motion planning p1
 Vehicle dynamics
 Vehicle dynamics
 F = ma
 thrust
 torque – moments
 Vehicle dynamics
 2D Vehicle Control
 Control Architecture
 P control (proportional control)
 PD control (proportional, derivative control)
 Control Architecture
 3D Vehicle Control
 Control Architecture
 PD cascaded control (proportional, derivative, nested control loops)
 PID architecture
 (proportional, derivative, integral multirate nested control loops
 Configuration space
 Geodetic / World Frame
 long, lat, alt.
 NED / Body / Local Frame
 north, east, down(pos)
 Geodetic / World Frame

 Linear algebra
 Integration
 Derivatives
 Linear algebra
 Linearization
 Matrix math
 Rgb (rotation matrices) global frame to body frame
 Euler angles
 V3F function
 Quaternion transformations
 Control Architecture
 State estimation

 UAVs (unmanned aerial vehicles), UASs (unmanned aerial systems)
 Geodetic Frame or World Frame
 ECEF – Earth Centered / Earth Fixed
 NED (North, East, Down)
 Local or Body Frame used in Aeronautics
 Euler Angles
 Gimbal lock
 Rotation Matrices
 Quaternions
 Motions as Transformations
 Configuration Space