for path planning using cnns
play

for Path Planning using CNNs Eike Rehder, Jannik Quehl and Christoph - PowerPoint PPT Presentation

Jun 30, 2023 •240 likes •807 views

Driving Like a Human: Imitation Learning for Path Planning using CNNs Eike Rehder, Jannik Quehl and Christoph Stiller eike.rehder@daimler.com IROS 2017 | 24.09.2017 KIT The Research University in the Helmholtz Association www.kit.edu

  1. Driving Like a Human: Imitation Learning for Path Planning using CNNs Eike Rehder, Jannik Quehl and Christoph Stiller eike.rehder@daimler.com IROS 2017 | 24.09.2017 KIT – The Research University in the Helmholtz Association www.kit.edu

  2. Introduction: Path Planning 2 Eike Rehder | IROS 2017 | 24.09.2017

  3. Short Review: Dijkstra‘s Algorithm 3 Eike Rehder | IROS 2017 | 24.09.2017

  4. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: Start Goal 4 Eike Rehder | IROS 2017 | 24.09.2017

  5. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 ? 4 ? 2 0 4 ? Start 4 Goal 2 2 ? ? 5 Assign edge costs, node costs, Start = 0 5 Eike Rehder | IROS 2017 | 24.09.2017

  6. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 ? 2 0 4 ? Start 4 Goal 2 2 ? 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs 6 Eike Rehder | IROS 2017 | 24.09.2017

  7. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 6 2 0 4 ? Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node 7 Eike Rehder | IROS 2017 | 24.09.2017

  8. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 6 2 0 4 ? Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node 8 Eike Rehder | IROS 2017 | 24.09.2017

  9. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 6 2 0 4 ? Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node 9 Eike Rehder | IROS 2017 | 24.09.2017

  10. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 5 2 0 4 ? Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node Re-assign minimum cost 10 Eike Rehder | IROS 2017 | 24.09.2017

  11. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 5 2 0 4 7 Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node Re-assign minimum cost 11 Eike Rehder | IROS 2017 | 24.09.2017

  12. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 1 4 4 5 2 0 4 7 Start 4 Goal 2 2 7 2 5 Assign edge costs, node costs, Start = 0 Propagate and sum costs Expand cheapest node Re-assign minimum cost Trace back shortest path 12 Eike Rehder | IROS 2017 | 24.09.2017

  13. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: Goal Start 13 Eike Rehder | IROS 2017 | 24.09.2017

  14. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: Graph Edges 14 Eike Rehder | IROS 2017 | 24.09.2017

  15. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: Obstacle 15 Eike Rehder | IROS 2017 | 24.09.2017

  16. Short Review: Dijkstra‘s Algorithm Find shortest path from start to goal: 16 Eike Rehder | IROS 2017 | 24.09.2017

  17. Shortest Path with a CNN 17 Eike Rehder | IROS 2017 | 24.09.2017

  18. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 18 Eike Rehder | IROS 2017 | 24.09.2017

  19. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate 19 Eike Rehder | IROS 2017 | 24.09.2017

  20. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate = 20 Eike Rehder | IROS 2017 | 24.09.2017

  21. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate and sum costs = + 1 = 21 Eike Rehder | IROS 2017 | 24.09.2017

  22. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate and sum costs +1 +1 +1 = +1 +0 22 Eike Rehder | IROS 2017 | 24.09.2017

  23. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate and sum costs = 23 Eike Rehder | IROS 2017 | 24.09.2017

  24. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate and sum costs Re-assign minimum cost min = 24 Eike Rehder | IROS 2017 | 24.09.2017

  25. Finding the Shortest Path with a CNN Assign edge costs, node costs, Start = 0 Propagate and sum costs Re-assign minimum cost min = 25 Eike Rehder | IROS 2017 | 24.09.2017

  26. Finding the Shortest Path with a CNN Replace min + pool Transition Filters Transition Cost Cost Updated Non-Zero Cost per Cost Padding (!) Action “Reinforcement learning via recurrent convolutional neural networks”, arXiv:1701.02392 26 Eike Rehder | IROS 2017 | 24.09.2017

  27. Finding the Shortest Path with a CNN Replace min + pool Transition Filters Transition Cost Cost Updated Non-Zero Cost per Cost Padding (!) Action Argmin of this layer is transition policy 27 Eike Rehder | IROS 2017 | 24.09.2017

  28. Evaluating the Shortest Path with a CNN + argmin * Flipped Transition Filters Current Transition Transition Next Destination State Policy Selection State State 28 Eike Rehder | IROS 2017 | 24.09.2017

  29. Example: Simple Path Planning 29 Eike Rehder | IROS 2017 | 24.09.2017

  30. Example: Path Planning Find shortest path from start to goal: Goal Obstacle Start 30 Eike Rehder | IROS 2017 | 24.09.2017

  31. Example: Path Planning Nine possible transition filters Cost is the traversed distance +√2 +√2 +1 +1 +0 +1 +√2 +√2 +1 31 Eike Rehder | IROS 2017 | 24.09.2017

  32. Example: Path Planning Cost Model Additive layer High cost where obstacle is located 32 Eike Rehder | IROS 2017 | 24.09.2017

  33. Example: Path Planning Cost Map State Transition Map 33 Eike Rehder | IROS 2017 | 24.09.2017

  34. Finding the Shortest Path with a CNN If you use Dijkstra: Graph traversal with known transitions is faster States can be updated selectively Visited nodes will never be touched again Why would you do it then? 34 Eike Rehder | IROS 2017 | 24.09.2017

  35. Driving Like a Human: Imitation Learning 35 Eike Rehder | IROS 2017 | 24.09.2017

  36. Imitation Learning Intersection in Karlsruhe Arial view: Google Maps 36 Eike Rehder | IROS 2017 | 24.09.2017

  37. Imitation Learning Recorded trajectories Teach a network to imitate human behavior Intersection in Karlsruhe Arial view: Google Maps 37 Eike Rehder | IROS 2017 | 24.09.2017

  38. Imitation Learning Replace min + pool Transition Filters Transition Cost Cost Updated Non-Zero Cost per Cost Padding (!) Action 38 Eike Rehder | IROS 2017 | 24.09.2017

  39. Imitation Learning Replace min + pool Transition Filters Transition Cost Cost Updated Non-Zero Cost per Cost Padding (!) Action Fill in the whole bunch of CNN techniques 39 Eike Rehder | IROS 2017 | 24.09.2017

  40. Example II: Imitation Learning + In our case: FC-ResNet operating on the arial view 40 Eike Rehder | IROS 2017 | 24.09.2017

  41. Imitation Learning Path driven by human 41 Eike Rehder | IROS 2017 | 24.09.2017

  42. Imitation Learning Path driven by human Cost map from arial image 42 Eike Rehder | IROS 2017 | 24.09.2017

  43. Imitation Learning Path driven by human Cost map after planning 43 Eike Rehder | IROS 2017 | 24.09.2017

  44. Imitation Learning Path planned by network Cost map after planning 44 Eike Rehder | IROS 2017 | 24.09.2017

  45. Imitation Learning Path planned by network Cost map after planning Path driven by human 45 Eike Rehder | IROS 2017 | 24.09.2017

  46. Imitation Learning Path planned by network Path driven by human 46 Eike Rehder | IROS 2017 | 24.09.2017

  47. Outlook: Prediction and Cooperation 47 Eike Rehder | IROS 2017 | 24.09.2017

  48. Outlook: Pedestrian Prediction Camera image Road Sidewalk Obstacles Semantic map and top view Teach a network to predict human motion by planning 48 Eike Rehder | IROS 2017 | 24.09.2017

  49. Outlook: Pedestrian Prediction Road Sidewalk Obstacles Pedestrian Crop of map centered around pedestrian “Pedestrian Prediction by Planning using Deep Neural Networks”, arXiv:1706.05904 49 Eike Rehder | IROS 2017 | 24.09.2017

  50. Outlook: Pedestrian Prediction Road Sidewalk Obstacles Pedestrian Predict destination for planning “Pedestrian Prediction by Planning using Deep Neural Networks”, arXiv:1706.05904 50 Eike Rehder | IROS 2017 | 24.09.2017

  51. Outlook: Pedestrian Prediction Road Sidewalk Obstacles Pedestrian Predicted with Net “Pedestrian Prediction by Planning using Deep Neural Networks”, arXiv:1706.05904 51 Eike Rehder | IROS 2017 | 24.09.2017

  52. Outlook: Cooperative Planning “ Cooperative Motion Planning for Non-Holonomic Agents Teach a network resolve conflicts with Value Iteration Networks ”, arXiv:1706.05904 52 Eike Rehder | IROS 2017 | 24.09.2017

  53. Summary 53 Eike Rehder | IROS 2017 | 24.09.2017

  54. Summary Planning Net… … for imitation … for prediction … for cooperation 54 Eike Rehder | IROS 2017 | 24.09.2017

  55. The People Jannik Quehl Trajectory Data Maximilian Naumann Cooperative Planning Florian Wirth Destination Prediction 55 Eike Rehder | IROS 2017 | 24.09.2017

Recommend

Maxim Likhachev 1 Motion/Path Planning Uncertainty and Planning Uncertainty can be in: -

Maxim Likhachev 1 Motion/Path Planning Uncertainty and Planning Uncertainty can be in: -

Motion/Path Planning Task: find a feasible (and cost-minimal) path/motion from CSE-571 the current configuration of the robot to its goal configuration (or one of its goal configurations) Deterministic Path Planning in Robotics Two

408 views • 11 slides
Geometric Path Planning McGill COMP 765 Oct 12 th , 2017 The Motion Planning Problem

Geometric Path Planning McGill COMP 765 Oct 12 th , 2017 The Motion Planning Problem

Geometric Path Planning McGill COMP 765 Oct 12 th , 2017 The Motion Planning Problem Intuition: Find a safe path/trajectory from start to goal More precisely: A path is a series of robot configurations (e.g., joint angles, points in

1.02k views • 61 slides
Maxim Likhachev 1 Planning via Cell Decomposition Planning via Cell Decomposition Graph

Maxim Likhachev 1 Planning via Cell Decomposition Planning via Cell Decomposition Graph

Motion/Path Planning Motion/Path Planning Examples (of what is usually referred to as path planning): Examples (of what is usually referred to as motion planning): Planned motion for a 6DOF robot arm CSE 473:Planning in Robotics (slides

573 views • 3 slides
CS 603 - Path Planning Rod Grupen 4/23/20 Robotics 1 4/23/20 Robotics 2 Why Motion

CS 603 - Path Planning Rod Grupen 4/23/20 Robotics 1 4/23/20 Robotics 2 Why Motion

Why Path Planning? CS 603 - Path Planning Rod Grupen 4/23/20 Robotics 1 4/23/20 Robotics 2 Why Motion Planning? Origins of Motion Planning GE Virtual Prototyping Character Animation T. Lozano-Prez and M.A. Wesley: Structural

310 views • 14 slides
Theta* Efficient Any-Angle Path Planning Kenny Daniel and Alex Nash March 26, 2007 Path

Theta* Efficient Any-Angle Path Planning Kenny Daniel and Alex Nash March 26, 2007 Path

Theta* Efficient Any-Angle Path Planning Kenny Daniel and Alex Nash March 26, 2007 Path Planning Overview Uses for Path Planning Navigation Games Mars Rovers Existing Algorithms Dijkstra A* D* lite

488 views • 25 slides
CMU-Q 15-381 Lecture 4: Path Planning Teacher: Gianni A. Di Caro A PPLICATION : M OTION P

CMU-Q 15-381 Lecture 4: Path Planning Teacher: Gianni A. Di Caro A PPLICATION : M OTION P

CMU-Q 15-381 Lecture 4: Path Planning Teacher: Gianni A. Di Caro A PPLICATION : M OTION P LANNING Path planning: computing a continuous sequence (a path) of configurations (states) between an initial configuration (start) and a final

367 views • 32 slides
CSE-571 Deterministic Path Planning in Robotics Courtesy of Maxim Likhachev University of

CSE-571 Deterministic Path Planning in Robotics Courtesy of Maxim Likhachev University of

CSE-571 Deterministic Path Planning in Robotics Courtesy of Maxim Likhachev University of Pennsylvania Motion/Path Planning Task: find a feasible (and cost-minimal) path/motion from the current configuration of the robot to its goal

602 views • 43 slides
Task and Path Planning for Multi-Agent Pickup and Delivery Minghua Liu Hang Ma Jiaoyang Li

Task and Path Planning for Multi-Agent Pickup and Delivery Minghua Liu Hang Ma Jiaoyang Li

. Background . . . . . . . . . . Task Assignment . Prioritized Path Planning Hybrid Path Planning Experimental Results Task and Path Planning for Multi-Agent Pickup and Delivery Minghua Liu Hang Ma Jiaoyang Li Sven Koenig

472 views • 43 slides
Gated Path Planning Networks Lisa Lee Machine Learning Department Carnegie Mellon University

Gated Path Planning Networks Lisa Lee Machine Learning Department Carnegie Mellon University

Gated Path Planning Networks Lisa Lee Machine Learning Department Carnegie Mellon University Joint work with Emilio Parisotto, Devendra Chaplot, Eric Xing, & Ruslan Salakhutdinov ICML 2018 Path Planning Gated Path Planning Networks

280 views • 26 slides
Integrating Dynamics into NTU, Singapore Industrial Motion Planning Path planning problem Find

Integrating Dynamics into NTU, Singapore Industrial Motion Planning Path planning problem Find

Q.-C. Pham Integrating Dynamics into NTU, Singapore Industrial Motion Planning Path planning problem Find a collision-free path between q s t a r t and q g o a l From academic breakthroughs... Configuration space formulation (Lozano-Perez

514 views • 27 slides
Humanoid Robotics Path Planning and Walking Maren Bennewitz 1 Motivation Given the

Humanoid Robotics Path Planning and Walking Maren Bennewitz 1 Motivation Given the

Humanoid Robotics Path Planning and Walking Maren Bennewitz 1 Motivation Given the robots pose in a model of the environment Compute a path to a target location First: 2D path in a 2D grid map representation of the

768 views • 56 slides
Understanding Geometry of Encoder-Decoder CNNs (E-D CNNs) Jong Chul Ye & Woon Kyoung Sung

Understanding Geometry of Encoder-Decoder CNNs (E-D CNNs) Jong Chul Ye & Woon Kyoung Sung

Understanding Geometry of Encoder-Decoder CNNs (E-D CNNs) Jong Chul Ye & Woon Kyoung Sung BISPL - BioImaging, Signal Processing and Learning Lab. Dept. Bio & Brain Engineering Dept. of Mathematical Sciences KAIST, Korea E-D CNN

911 views • 26 slides
A Path Planning Approach for Computing A Path Planning Approach for Computing Large- -Amplitude

A Path Planning Approach for Computing A Path Planning Approach for Computing Large- -Amplitude

A Path Planning Approach for Computing A Path Planning Approach for Computing Large- -Amplitude Motions of Flexible Molecules Amplitude Motions of Flexible Molecules Large Juan Corts & Thierry Simon MOVIE Workshop, Toulouse,

402 views • 22 slides
CS 403 - Path Planning Roderic A. Grupen 4/2/19 Robotics 1 4/2/19 Robotics 2 Why

CS 403 - Path Planning Roderic A. Grupen 4/2/19 Robotics 1 4/2/19 Robotics 2 Why

Why Motion Planning? CS 403 - Path Planning Roderic A. Grupen 4/2/19 Robotics 1 4/2/19 Robotics 2 Why Motion Planning? Summary GE Control Virtual Prototyping Character Animation Kinematics Structural Molecular Biology

193 views • 16 slides
Table of Contents Convolutional Neural Nets (CNNs) 1 Deep Q Learning 2 Lecture 6: CNNs and Deep

Table of Contents Convolutional Neural Nets (CNNs) 1 Deep Q Learning 2 Lecture 6: CNNs and Deep

Lecture 6: CNNs and Deep Q Learning 2 Emma Brunskill CS234 Reinforcement Learning. Winter 2018 2 With many slides for DQN from David Silver and Ruslan Salakhutdinov and some vision slides from Gianni Di Caro and images from Stanford CS231n,

850 views • 67 slides
Table of Contents Convolutional Neural Nets (CNNs) 1 Deep Q Learning 2 Lecture 6: CNNs and Deep

Table of Contents Convolutional Neural Nets (CNNs) 1 Deep Q Learning 2 Lecture 6: CNNs and Deep

Lecture 6: CNNs and Deep Q Learning 1 Emma Brunskill CS234 Reinforcement Learning. Winter 2019 1 With many slides for DQN from David Silver and Ruslan Salakhutdinov and some vision slides from Gianni Di Caro and images from Stanford CS231n,

1.17k views • 68 slides
11/27/2006 Massachusetts Institute of Technology Context Optimal path planning for dynamic

11/27/2006 Massachusetts Institute of Technology Context Optimal path planning for dynamic

11/27/2006 Massachusetts Institute of Technology Context Optimal path planning for dynamic systems Optimal, Robust Path Planning What is best sequence of control inputs that takes system state from the start to goal? a

171 views • 5 slides
Bounded Suboptimal Path Planning with Compressed Path Databases Shizhe Zhao 1 , Mattia Chiari 2 ,

Bounded Suboptimal Path Planning with Compressed Path Databases Shizhe Zhao 1 , Mattia Chiari 2 ,

Bounded Suboptimal Path Planning with Compressed Path Databases Shizhe Zhao 1 , Mattia Chiari 2 , Adi Botea 3 , Alfonso E. Gerevini 2 , Daniel Harabor 1 , Alessandro Saetti 2 , Peter J. Stuckey 1 1 Monash University, 2 University of Brescia, 3

227 views • 19 slides
RRT-Connect: An Efficient Approach to Single-Query Path Planning James J. Kuffner, Jr. Steven M.

RRT-Connect: An Efficient Approach to Single-Query Path Planning James J. Kuffner, Jr. Steven M.

RRT-Connect: An Efficient Approach to Single-Query Path Planning James J. Kuffner, Jr. Steven M. LaValle ICRA 2000 Presented by Manel Baradad 6.882 Embodied Intelligence Spring 2020 Problem Single-query path planning, from start to goal. In

305 views • 26 slides
P.A.T.H Planning tool Presentation to the Whnau Ora Hui Sharing the Learning Auckland, 15

P.A.T.H Planning tool Presentation to the Whnau Ora Hui Sharing the Learning Auckland, 15

P.A.T.H Planning tool Presentation to the Whnau Ora Hui Sharing the Learning Auckland, 15 16 August 2011 Kataraina Pipi and Mariao Hohaia (PATH Facilitators) Overview What is the PATH? Experience of the PATH in Aotearoa

524 views • 24 slides
Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are

Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are

IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS Geirhos et al. (2019) Introduction ImageNet classifjcation with CNNs Which image cues are learned How infmuential they are

221 views • 18 slides
Path Planning Jan Faigl Department of Computer Science Faculty of Electrical Engineering Czech

Path Planning Jan Faigl Department of Computer Science Faculty of Electrical Engineering Czech

Path Planning Jan Faigl Department of Computer Science Faculty of Electrical Engineering Czech Technical University in Prague Lecture 03 B4M36UIR Artificial Intelligence in Robotics Jan Faigl, 2020 B4M36UIR Lecture 03: Path Planning

1.65k views • 140 slides
Artificial Intelligence: Methods and applications Lecture 6: Path planning Ola Ringdahl Ume

Artificial Intelligence: Methods and applications Lecture 6: Path planning Ola Ringdahl Ume

Artificial Intelligence: Methods and applications Lecture 6: Path planning Ola Ringdahl Ume University November 21, 2014 Navigation Four questions: Where am I going? Mission planning (human or planner) What is the best way to

992 views • 49 slides
Planning and Optimization E1. Critical Path Heuristics: h m Malte Helmert and Gabriele R oger

Planning and Optimization E1. Critical Path Heuristics: h m Malte Helmert and Gabriele R oger

Planning and Optimization E1. Critical Path Heuristics: h m Malte Helmert and Gabriele R oger Universit at Basel November 20, 2017 Set Representation Perfect Regression Heuristic Critical Path Heuristics Computation Summary Content of

834 views • 38 slides

More recommend