1 / 88

Real Time Motion Planning and Safe Navigation in Dynamic Environments*

Real Time Motion Planning and Safe Navigation in Dynamic Environments*. Kadir F. Uyanik CENG585 Fundamentals of Autonomous Robotics 14.01.2011. * Based on: Bruce J. R., Real-Time Motion Planning and Safe Navigation in Dynamic Multi-Robot Environments , PhD. Thesis, 2006.

kasi
Download Presentation

Real Time Motion Planning and Safe Navigation in Dynamic Environments*

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. Real Time Motion Planning and Safe Navigation in Dynamic Environments* Kadir F. Uyanik CENG585 Fundamentals of Autonomous Robotics 14.01.2011 * Based on: Bruce J. R., Real-Time Motion Planning and Safe Navigation in Dynamic Multi-Robot Environments, PhD. Thesis, 2006 * Some of the slides are adapted from James Bruce’s PhD. Defense presentation.

  2. Thesis Goal • Enabling a multi agent system carry out navigation calculations within tight time constraints • Making robots navigate robustly and operate safely without collisions

  3. Outline • Randomized Approaches • RRT, RRT-Connect • ERRT, ERRT-MultiConnect • ERRT vs Visibility Graphs • Novelty up-to now • From Kinematic Planning to Dynamic Planning • Dynamic Window Method • Dynamic Safety Search • Conclusion • Introduction • A classification in robotic systems • Robot soccer • Small Size League (SSL) system • Navigation System for SSL Robots • Planning motions • Planning in a changing world • Problem Definition • Common Approaches • Grid Based • Visibility Graph • Randomized Sampling Based • Planning Challenges

  4. Introduction

  5. Introduction

  6. Introduction

  7. Introduction

  8. Introduction

  9. Introduction

  10. Introduction

  11. Introduction

  12. Introduction

  13. Introduction

  14. Introduction

  15. Introduction

  16. Introduction

  17. IntroductionSoccer Playing Robots Two main worldwide competitions/organizations:

  18. IntroductionSoccer Playing Robots Two main worldwide competitions/organizations: • FIRA Mirosot: Micro-Robot World Cup Soccer Tournament. Organized by Federation of International Robot-Soccer Association since 1996.

  19. IntroductionSoccer Playing Robots Two main worldwide competitions/organizations: • FIRA Mirosot: Micro-Robot World Cup Soccer Tournament. Organized by Federation of International Robot-Soccer Association since 1996. • Robocup: Robot World Cup, largest international robotics competition. Organized (officially) since 1997. This year in Istanbul/Turkey (June 4-10, 2011) Several categories: Soccer, rescue, @home, junior Soccer includes various leagues: humanoid, middle size, small size, standard platform, simulation

  20. IntroductionSmall Size League Robot Soccer System

  21. IntroductionSmall Size Soccer League

  22. Navigation System for SSL • Plan quickly before planned decisions become obsolete • Agents act parallel in multi-robot domains; unpredictable dynamics can arise, • Other team’s robots move very fast and world changes quickly.

  23. Navigation System for SSL • Plan quickly before planned decisions become obsolete • Agents act parallel in multi-robot domains; unpredictable dynamics can arise, • Other team’s robots move very fast and world changes quickly. • Navigate robustly, don’t crash other robots , stay in the field

  24. Navigation System for SSLPlanning motions • Motion Planning is about finding trajectories to satisfy a goal criteria starting from an initial-configuration to an end-configuration.

  25. Navigation System for SSLPlanning motions • Motion Planning is about finding trajectories to satisfy a goal criteria starting from an initial-configuration to an end-configuration. • Two main requirements are • Model of the environment or the world state is known to some degree • Model of the results of actions that create certain effect in the world

  26. Navigation System for SSLPlanning motions • Motion Planning is about finding trajectories to satisfy a goal criteria starting from an initial-configuration to an end-configuration. • Two main requirements are • Model of the environment or the world state is known to some degree • Model of the results of actions that create certain effect in the world • This knowledge enables robot, in a way, to simulate it’s actions in mind and predict the output w/o actually executing them.

  27. Navigation System for SSLPlanning in a changing world • It is a key issue in a multi-agent systems

  28. Navigation System for SSLPlanning in a changing world • It is a key issue in a multi-agent systems • Agent dynamics are the limitations due to the kinodynamicconstraints of the robots

  29. Navigation System for SSLPlanning in a changing world • It is a key issue in a multi-agent systems • Agent dynamics are the limitations due to the kinodynamic constraints of the robots • Domain dynamics includes environmental changes (due to other robots or physical laws) and changes in goal specification (due to higher level task oriented behaviors)

  30. Problem Definition A : agent q : robot configuration Cfree : obstacle free configuration space T(s) : continuous function, mapping s ͼ [0,1] to a configuration in C. Rj(t) : area covered by all robots except j S’(t) : boolean safety function (true if no two robots overlap) Given : A, Cfree , qinit , qgoal ; Find : a path T(s) which is valid, feasible, and a solution. For a safe navigation:

  31. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  32. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  33. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  34. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  35. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  36. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan

  37. Common Approaches • A generic motion (re)planning algorithm: • Map initial and goal locations to C-space representation • Update environment model with new information • Update C-space representation graph, or roadmap • Search roadmap for a path between initial and goal locations • Extract path vertices and edges as plan • Replanning can be done in two ways • Unconditional replanning: replan each time before deciding on an action • Conditional replanning: Plan once, monitor the environment and execution of plan to determine if it succeeds or fails. If fails, replan and continue execution.

  38. Common Approaches • The planners used in SSL-like domains are based on: • Grids • Create a grid overlay of vertices covering the environment • Connect grid neighbors with edges if free • Search for shortest (least cost) path • Non-optimal and complete • Visibility Graph • Place vertices at critical points around each obstacle • Add edges between every pair of critical points if free • Optimal and complete • Randomized • No need for grids or list of obstacle points; discover Cfree through collision checks • Sample environment randomly to model C-space • Search until tree reaches a goal • Non-optimal and probabilistically complete

  39. Planning Challenges It is all about Time vs. Problem complexity Dashed and straight curved lines indicates the corresponding hypothetical algorithm performance

  40. Outline • Randomized Approaches • RRT, RRT-Connect • ERRT, ERRT-MultiConnect • ERRT vs Visibility Graphs • Novelty up-to now • From Kinematic Planning to Dynamic Planning • Dynamic Window Method • Dynamic Safety Search • Conclusion • Introduction • A classification in robotic systems • Robot soccer • Small Size League (SSL) system • Navigation System for SSL Robots • Planning motions • Planning in a changing world • Problem Definition • Common Approaches • Grid Based • Visibility Graph • Randomized Sampling Based • Planning Challenges

  41. Outline • Randomized Approaches • RRT, RRT-Connect • ERRT, ERRT-MultiConnect • ERRT vs Visibility Graphs • Novelty up-to now • From Kinematic Planning to Dynamic Planning • Dynamic Window Method • Dynamic Safety Search • Conclusion • Introduction • A classification in robotic systems • Robot soccer • Small Size League (SSL) system • Navigation System for SSL Robots • Planning motions • Planning in a changing world • Problem Definition • Common Approaches • Grid Based • Visibility Graph • Randomized Sampling Based • Planning Challenges

  42. Randomized ApproachesRapidly exploring random trees(RRT) • Start with the initial state as the root of a tree • Pick a random state in anywhere or in the direction of the target • Find the closest node in the current tree • Extend that node toward the target if possible

  43. Randomized ApproachesRapidly exploring random trees(RRT) • Start with the initial state as the root of a tree • Pick a random state in anywhere or in the direction of the target • Find the closest node in the current tree • Extend that node toward the target if possible

  44. Randomized ApproachesRapidly exploring random trees(RRT) • Start with the initial state as the root of a tree • Pick a random state in anywhere or in the direction of the target • Find the closest node in the current tree • Extend that node toward the target if possible

  45. Randomized ApproachesRapidly exploring random trees(RRT) • Start with the initial state as the root of a tree • Pick a random state in anywhere or in the direction of the target • Find the closest node in the current tree • Extend that node toward the target if possible

  46. Randomized ApproachesRapidly exploring random trees(RRT)

  47. Randomized ApproachesRapidly exploring random trees(RRT) • Don't add extensions which would hit obstacles • Resulting tree contains only valid paths

  48. Randomized ApproachesRapidly exploring random trees(RRT) • Search until goal is reached • Backtrack the tree

  49. Randomized ApproachesRapidly exploring random trees(RRT)

  50. Randomized ApproachesExecution Extended RRT (ERRT) • Waypoints • Idea: Previously successful plans can guide new search • Biases can be encoded in the target distribution • The waypoint cache • Whenever a plan is found, store nodes in a fixed-size bin with random replacement • During random target point selection, sometimes choose a waypoint from the cache

More Related