1 / 27

Ch. 5: Path Planning

Ch. 5: Path Planning. Updates. Lab #3 next week Prelab distributed today Lab #2 writeup due 3/22 Talk on Friday: Brian Scassellati, Yale MD G125, 3:00-4:00 “Social Robots and Human Social Development” Website update. Path planning overview.

irish
Download Presentation

Ch. 5: Path Planning

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Ch. 5: Path Planning ES159/259

  2. Updates • Lab #3 next week • Prelab distributed today • Lab #2 writeup due 3/22 • Talk on Friday: • Brian Scassellati, Yale • MD G125, 3:00-4:00 • “Social Robots and Human Social Development” • Website update ES159/259

  3. Path planning overview • Want to find a path from an initial position to a final position • Both are defined as vectors in the configuration space • Initial configuration is qs and final configuration is qf • Instead of completely defining QO or Qfree, we develop a sequence of discrete configurations that drive the configuration from qs to qf and avoid obstacles • Definition: • g: continuous mapping from qs to qf • Such that: ES159/259

  4. Potential fields • To develop the mapping, we incrementally explore Qfree • Consider the manipulator (statically) as a point in the configuration space • The manipulator is subject to a potential field • Attractive in the case of the goal configuration • Repulsive in the case of an obstacle • Therefore we want to find the minimum potential • Hopefully, this will be at the global minimum configuration qf ES159/259

  5. Gradient descent • In order to find minima of U, take the negative gradient: • Where t is a vector of joint forces/torques • in the configuration space • However, it is difficult to construct the potential field in the configuration space • Due to difficulty in describing the obstacles in the configuration space • Due to potential collisions with the links of the manipulator • Therefore we define the potential field not in the configuration space, but in the workspace • Thus the negative gradient will be a vector of forces in the workspace ES159/259

  6. Hybrid attractive field • Combine the conic well potential and parabolic well potential fields • If the ith frame is close to the workspace goal, use the parabolic well • If the ith frame is far from the workspace goal, use the conic well • The distance d defines the distance from the goal that causes a transition from a conic to parabolic potential • Since this is continuous everywhere, the workspace force is defined everywhere ES159/259

  7. Hybrid attractive field • Taking the gradient gives the workspace attractive force ES159/259

  8. The repulsive field • Prevent collisions by creating a repulsive force in the workspace • Again, create forces that act on the origins of the n DH coordinate frames • These forces should: • Repel the robot from obstacles • Do nothing of the robot is far away from obstacles • First, define a radius of influence, r0, of an obstacle • The object will not repel oi if the distance is greater than r0 • Where r(oi(q)) is the shortest distance between oi and any workspace obstacle ES159/259

  9. The repulsive field • So we can write this force as: • Finally, we need to define the gradient of the distance function r(oi(q)) • i.e. the gradient of r(x) at x = oi(q) • This defines the direction of the repulsive force (it is a unit vector) • Where b is the point on the obstacle that is closest to oi ES159/259

  10. The relation between workspace forces and joint torques • When we discussed velocity kinematics we said that we could use the Jacobian to describe the mapping from joint forces and torques to end effector forces and torques: • Jv is the velocity portion of the Jacobian (since we have only discussed workspace forces, not torques) • The Jacobian must be derived for each origin oi, denoted Joi ES159/259

  11. Ex: two-link planar manipulator • Consider the previous examples with an obstacle exerting a repulsive force on o2 • Find the attractive and repulsive forces on o1 and o2 • First determine the Jacobians for the two coordinate frames Initial and goal configurations Obstacle location ES159/259

  12. Ex: two-link planar manipulator • For o1: • For o2: • From the previous examples we have: • Workspace attractive force on o1: • Workspace attractive force on o2: • Workspace repulsive force on o1: • Workspace repulsive force on o2: ES159/259

  13. Ex: two-link planar manipulator • To determine the joint torques, take the transpose of the Jacobians at the initial configuration ES159/259

  14. Ex: planar mobile robot • Using potential fields to control a mobile robot in the plane is similar to controlling the motion of the end effector in the workspace • Configuration determined by the generalized coordinates q = (x,y,q) • The point a has coordinates (ax,ay) such that: • Therefore the Jacobian for point a is: ES159/259

  15. Ex: planar mobile robot • Multiplying the transpose of the Jacobian by the workspace force gives the forces and torques in the configuration space: ES159/259

  16. Composing workspace forces • The total joint torques acting on a manipulator is the sum of the torques from all attractive and repulsive potentials: • If we add workspace forces before transforming into the configuration space, we will have incorrect torques in the configuration space • For example, consider two forces in the workspace such that F1 + F2 = 0 • Adding these in the workspace will ignore the torque about the center ES159/259

  17. Ex: two-link planar manipulator • Consider again the two-link manipulator with a goal position and an obstacle near o2 • The total joint torque, due to these two potential fields is: Initial and goal configurations Obstacle location ES159/259

  18. Gradient descent • Now that we can formulate the total torques acting on the joints in the configuration space due to the artificial potentials, we can formulate a path planning algorithm • First, determine your initial configuration • Second, given a desired point in the workspace, calculate the final configuration using the inverse kinematics • Use this to create an attractive potential field • Locate obstacles in the workspace • Create a repulsive potential field • Sum the joint torques in the configuration space • Use gradient descent to reach your target configuration ES159/259

  19. Algorithm: Notation: qi: configuration at the ith iteration e: region of convergence zi:influence of attractive potential for oi. Can be different for each oi ai:defines step size of ith iteration hi:influence of repulsive potential for oi. Can be different for each oi r0:distance of influence for obstacles. Can be defined differently for different obstacles Gradient descent ES159/259

  20. Local minima • In the absence of obstacles, the gradient descent will always converge to the global minimum (qf) • With obstacles, by proper choice of ai, this will always converge to some minima • May not be the global minimum • Probabilistic methods to escape local minima ES159/259

  21. Instead we modify the gradient descent algorithm to add a random excitation in case we are stuck in a local minima We are stuck in a local minima if successive iterations result in minimal changes in the configuration If so, perform a random walk to get out The random walk is defined by adding a uniformly distributed variable to each joint parameter Local minima ES159/259

  22. Random walk • Take a number of random steps starting from a local minimum configuration • To do this, add a constant to each qi at each step in the random walk: • Each vi are fixed constants, and there is an equal probability of adding +vi or –vi • i.e. a uniform distribution • If this is carried out for t steps, the probability density function can be represented as a zero-mean Gaussian: ES159/259

  23. Alternate method: Probabilistic Roadmaps • Alternative to gradient descent • Randomly sample the configuration space (uniform) • Discard samples that lie within QO • This gives a rough idea of Qfree ES159/259

  24. Alternate method: Probabilistic Roadmaps • Now connect the nodes of the sampled configuration space • For each point, connect to the k nearest neighbors • To decide the nearest neighbors using a norm in either the workspace or configuration space • Connect using straight line segments that do not intersect obstacles ES159/259

  25. Alternate method: Probabilistic Roadmaps • Enhancement: It is clear that there are numerous connected segments, now connect as many of these as possible • Resample, keeping the previous segments • Can do this a number of times until Qfree is sufficiently covered ES159/259

  26. Alternate method: Probabilistic Roadmaps • Choose a path in the connected space • Given qs and qf, apply to the roadmap • Search the roadmap for a path from qs to qf ES159/259

  27. Next class… • Finish the discussion on path planning • Trajectory planning • Examples • Introduction to single variable control ES159/259

More Related