Robotic Motion Planning: Review C-Space and Start Potential Functions

Transcription

1 Robotic Motion Planning: Review C-Space and Start Potential Functions Robotics Institute Howie Choset

2 What if the robot is not a point? The Scout should probably not be modeled as a point... β α Nor should robots with extended linkages that may contact obstacles...

3 What is the position of the robot? Expand obstacle(s) Reduce robot not quite right...

4 Trace Boundary of Workspace Really the origin of a reference configuration Pick a reference point

5 Polygonal robot translating & rotating in 2-D workspace θ y x Reference configuration

6 Mapping from the workspace to the configuration space Reference configuration is horizontal configuration workspace configuration space The free space is generally an open set A free path is a mapping c:[0,1]æ Qfree A semifree path is a mapping c:[0,1]æ cl(qfree)

7 Minkowski sum The Minkowski sum of two sets P and Q, denoted by P Q, is defined as P+Q = { p+q p P, q Q } q Similarly, the Minkowski difference is defined as P Q = { p q p P, q Q } p

8 Minkowski sum of convex polygons The Minkowski sum of two convex polygons P and Q of m and n vertices respectively is a convex polygon P + Q of m + n vertices. The vertices of P + Q are the sums of vertices of P and Q.

9 Observation If P is an obstacle in the workspace and M is a moving object. Then the C-space obstacle corresponding to P is P M. M P O

10 Star Algorithm: Polygonal Obstacles e1 r1 r3 r2 e4 e2 e1 e3 r2 r3 e4 e2 e3 r1

11 Potential Functions Additive/Repulsive Local minima Navigation Functions

12 The Basic Idea A really simple idea: Suppose the goal is a point g R 2 Suppose the robot is a point r R 2 Think of a spring drawing the robot toward the goal and away from obstacles: Can also think of like and opposite charges

13 Another Idea Think of the goal as the bottom of a bowl The robot is at the rim of the bowl What will happen?

14 The General Idea Both the bowl and the spring analogies are ways of storing potential energy The robot moves to a lower energy configuration A potential function is a function U : R m R Energy is minimized by following the negative gradient of the potential energy function: We can now think of a vector field over the space of all q s... at every point in time, the robot looks at the vector at the point and goes in that direction

15 Attractive/Repulsive Potential Field U att is the attractive potential --- move to the goal U rep is the repulsive potential --- avoid obstacles

16 Conical Potential Artificial Potential Field Methods: Attractive Potential Quadratic Potential F att ( q) = U = k δ att goal ( q) ( q)

17 Artificial Potential Field Methods: Attractive Potential Combined Potential In some cases, it may be desirable to have distance functions that grow more slowly to avoid huge velocities far from the goal one idea is to use the quadratic potential near the goal (< d*) and the conic farther away One minor issue: what?

18 The Repulsive Potential

19 Repulsive Potential

20 Total Potential Function U ( q) = U att ( q) + U rep ( q) F( q) = U ( q) + =

21 Potential Fields

22 Gradient Descent A simple way to get to the bottom of a potential A critical point is a point x s.t. U(x) = 0 Equation is stationary at a critical point Max, min, saddle Stability?

23 The Hessian For a 1-d function, how do we know we are at a unique minimum (or maximum)? The Hessian is the m m matrix of second derivatives If the Hessian is nonsingular (Det(H) 0), the critical point is a unique point if H is positive definite (x^t H x > 0), a minimum if H is negative definite, a maximum if H is indefinite, a saddle point

24 Gradient Descent: q(0)=q start i = 0 while U(q(i)) 0 do q(i+1) = q(i) - α(i) U(q(i)) i=i+1 Gradient Descent

25 Gradient Descent: q(0)=q start i = 0 while U(q(i)) > ε do q(i+1) = q(i) - α(i) U(q(i)) i=i+1 Gradient Descent

26 Numerically Smoother Path

27 Single Object Distance

28 Compute Distance: Sensor Information

29 Computing Distance: Use a Grid use a discrete version of space and work from there The Brushfire algorithm is one way to do this need to define a grid on space need to define connectivity (4/8) obstacles start with a 1 in grid; free space is zero 4 8

30 Brushfire Algorithm Initially: create a queue L of pixels on the boundary of all obstacles While L pop the top element t of L if d(t) = 0, set d(t) to 1+min t N(t),d(t) 0 d(t ) Add all t N(t) with d(t)=0 to L (at the end) The result is a distance map d where each cell holds the minimum distance to an obstacle. The gradient of distance is easily found by taking differences with all neighboring cells.

31 Brushfire example

32 Potential Functions Question How do we know that we have only a single (global) minimum We have two choices: not guaranteed to be a global minimum: do something other than gradient descent (what?) make sure only one global minimum (a navigation function, which we ll see later).

33 Ming Lin s Voronoi Region Distance Explained this on the white board.. Then talked with Potential function guys