Incremental Path Planning

Size: px
Start display at page:

Download "Incremental Path Planning"

Transcription

1 Incremental Path Planning April 4, 206 Joe Leavitt, Ben Ayton, Jessica Noss, Erlend Harbitz, Jake Barnwell & Sam Pinto

2 References Koenig, S., & Likhachev, M. (2002, July). D* Lite. In AAAIIAAI (pp ). Koenig, S., Likhachev, M., & Furcy, D. (2004). Lifelong planning A. Artificial Intelligence, 55(), Stentz, A. (994, May). Optimal and efficient path planning for partially-kno n environments. In Proceedings of the IEEE International Conference on Robotics and Automation. Likhachev, M., Ferguson, D. I., Gordon, G. J., Stentz, A., & hrun, S. (200, June). Anytime Dynamic A*: An Anytime, Replanning Algorithm. In ICAPS (pp ). Hofmann, A., Fernandez, E., Helbert, J., Smith, S., & illiams,. (20). Reactive Integrated Motion Planning and Euecution. In Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence. 2

3 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 3

4 Motivation 4

5 Motivation 5

6 Motivation e obstacle detected Replan 6

7 Motivation e obstacle detected Replan 7

8 Motivation e obstacle detected Replan 8

9 Motivation e obstacle detected Replan 9

10 Motivation 0

11 Motivation

12 Motivation 2

13 Motivation 3

14 Motivation hange detected Replan 4

15 Motivation Environmental hange 5

16 Motivation hange detected Replan 6

17 Motivation hange detected Replan 7

18 Motivation 8

19 Motivation 9

20 Motivation 20

21 Motivation 2

22 Motivation 22

23 Motivation 23

24 Motivation 24

25 Motivation 25

26 Motivation 26

27 Motivation 27

28 Motivation RR Sertac Karaman. All rights reserved. This content is excluded from our Creative Commons license. For more information, see Watch the video on YouTube ( 28

29 Motivation After significant offline computation time... 29

30 Motivation RR * Sertac Karaman. All rights reserved. This content is excluded from our Creative Commons license. For more information, see 30

31 Problems Changing environmental conditions: o Obstacles o Utility or cost Sensor limitations: o Partial observability Computation time: o Complete, optimal replanning is slow o Stay put or move in wrong direction? 3

32 n Reuse data nrom previous search Incremental Search Methods 32

33 Incremental APSP Alg. Mean Std omp. Opt. hekhov (APSP) es es RR Prob. o RR connect Prob. o PRM Prob. Asym. Courtesy of Hofmann, Andreas et al. License: cc by-nc-sa. 33

34 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 34

35 Incremental Search Pernorm graph search Repeat: o Execute path/plan o Receive graph changes o Update previous search results 35 From Koenig, Likhachev, & Furcy (2004)

36 Revie of Graph Search Problem Input: graph search problem S = <gr, w, h, s start, s goal > directed graph: gr = <V, E> edge eighting function: w: s x s heuristic function: h: s x s goal start verteu: s V start goal verteu: s goal V Output: simple path P = <s, s 2, ", s goal > start s start (s,s ) start (s start,s 2 ) s (s,s goal ) h(s start,s goal ) (s 2,s goal ) s 2 h(s,s goal ) s goal h(s 2,s goal ) 36

37 Revie of Graph Search Problem g: cost-so-far (s start,s ) s h(s,s goal ) o g(s) = g(s predecessor ) + (s predecessor, s) (s,s goal ) h: heuristic value, cost-to-go s start h(s start,s goal ) s goal f(s) = g(s) + h(s) (s start,s 2 ) (s 2,s goal ) h(s 2,s goal ) s 2 37

38 Graph Representations Incremental algorithms generalize to any graph (usually non-negative edges). Grids used for ease of representation. 4-onnected Graph -onnected Graph Admissible Manhattan Distance mau(lu, ly) Heuristic: = lu + ly ( ith unit edge weights) 38

39 Relauation - Dikstras Algorithm

40 Relauation - Dikstras Algorithm

41 Relauation - Dikstras Algorithm

42 Relauation - Dikstras Algorithm

43 Relauation - Dikstras Algorithm

44 Relauation - Dikstras Algorithm

45 Relauation - Dikstras Algorithm

46 Relauation - Dikstras Algorithm

47 Relauation - Dikstras Algorithm

48 Relauation - Dikstras Algorithm

49 Relauation - Dikstras Algorithm

50 Relauation - Dikstras Algorithm

51 Relauation - Dikstras Algorithm

52 Relauation - Dikstras Algorithm

53 Relauation - Dikstras Algorithm

54 Relauation - Dikstras Algorithm

55 Ho to Reuse Previous Search Results Store optimal results: o shortest paths o g-values Find inconsistencies: o local consistency Make consistent: o relaxations (a) (b) (c) Courtesy of Elsevier, Inc., Used with permission. Source: Figures 3 and 4 in Keonig, Sven, M. Likhachev, and D. Furcy. "Lifelong Planning A*." In Artificial Intelligence, 55 (- 2): From Koenig, Likhachev, & 55 Furcy (2004)

56 Incremental Search Methods - Euamples General o o Incremental APSP Lifelong Planning A* (LPA*) o Dynamics Strictly eakly Superior Function - Fiued Point (DynamicsS SF-FP) o... Mobile Robots o D* o D* Lite emporal Planning o Incremental emporal onsistency (I ) Propositional Satisfiability o Incremental Unit Propagation 56

57 D* Incremental Path Planning Approach Heuristic Search (e.g. A*) Incremental + Search = (e.g. DynamicsS SF-FP) Efficient Incremental Path Planning (e.g. LPA*, D* Lite,...) Optimal & Efficient Fast Replanning Fast, Optimal Replanning 57

58 Initial Search Courtesy of Elsevier, Inc., Used with permission. Source: Figures 3 and 4 in Keonig, Sven, M. Likhachev, and D. Furcy. "Lifelong Planning A*." In Artificial Intelligence, 55 (- 2): From Koenig, Likhachev, & Furcy (2004)

59 Follo -on Search From Koenig, Likhachev, & Furcy (2004) Courtesy of Elsevier, Inc., Used with permission. Source: Figures 3 and 4 in Keonig, Sven, M. Likhachev, and D. Furcy. "Lifelong Planning A*." In Artificial Intelligence, 55 (- 2):

60 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 60

61 A* Reminder A* is best-first search from start to goal sorted by a cost f(s): f(s) = g(s) + h(s,s goal ) f(s) = total node cost g(s) = path cost to reach verteu from s start h(s,s goal ) = heuristic for cost to reach verteu s goal from verteu s 6

62 A*: hoosing et een Ties Introduce a ne notation: f(s) = ( f (s), f 2 (s) ) f(s) = ( g(s) + h(s,s goal ), g(s) ) 3 h = 2 h = s f(s ) = s start s goal f(s 2 ) = 2 s 2 h = 3 2 h = 62

63 A*: hoosing et een Ties Introduce a ne notation: f(s) = ( f (s), f 2 (s) ) f(s) = ( g(s) + h(s,s goal ), g(s) ) s start 3 h = 2 h = s f(s ) = (, 3 ) s goal f(s 2 ) = 2 s 2 h = 3 2 h = 63

64 A*: hoosing et een Ties Introduce a ne notation: f(s) = ( f (s), f 2 (s) ) f(s) = ( g(s) + h(s,s goal ), g(s) ) s start 3 h = 2 h = s f(s ) = (, 3 ) s goal f(s 2 ) = (, 2 ) 2 s 2 h = 3 2 h = 64

65 A*: hoosing et een Ties Introduce a ne notation: f(s) = ( f (s), f 2 (s) ) f(s) = ( g(s) + h(s,s goal ), g(s) ) 3 h = 2 h = s f(s ) = (, 3 ) s start s goal f(s 2 ) = (, 2 ) 2 s 2 h = 3 2 h = 65

66 Successors and Predecessors Successors of node s: Every node that can be reached from s, Succ(s) Predecessors of verteu s: Every node from which s can be reached, Pred(s) 66

67 Successors and Predecessors Successors of node s: Every node that can be reached from s, Succ(s) Predecessors of verteu s: Every node from which s can be reached, Pred(s) 67

68 Successors and Predecessors Successors of node s: Every node that can be reached from s, Succ(s) Predecessors of verteu s: Every node from which s can be reached, Pred(s) 68

69 Successors and Predecessors Successors of node s: Every node that can be reached from s, Succ(s) Predecessors of verteu s: Every node from which s can be reached, Pred(s) 69

70 hat is D* Lite Efficient repeated best-first search through a graph with changing edge weights as the graph is traversed. s start s goal s goal s start an be vie ed as replanning through relauation of path costs. 70

71 Reformulation: Minimize Recomputation When the start changes, the path cost from start to s is not preserved. ase : ase 2: s start 2 2 s s goal 2 s start 2 s s goal (s) = = 4 g g (s) t g 2 (s) g 2 (s) = 2 7

72 Reformulation: Minimize Recomputation Reformulate search from goal to start. Path cost from goal to s is preserved. ase : ase 2: s start 2 s goal s s s start s goal (s) = g g (s) = g 2 (s) g 2 (s) = 72

73 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s is consistent with neighbors and eupanded. start 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. b. Update source nodes of changed edges.. Repeat from 2. 73

74 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s start is consistent with neighbors and eupanded. start 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. Most computation b. Update source nodes of changed edges.. Repeat from 2. 74

75 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s start is consistent with neighbors and eupanded. 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. b. Update source nodes of changed edges.. Repeat from 2. Incremental component 75

76 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s start is consistent with neighbors and eupanded. 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. b. Update source nodes of changed edges.. Repeat from 2. 76

77 Eutracting a Path Given Path ost Move from s to the successor which gives s the lo est path cost. start start s argmin (c(s,s) + g(s)) start s E Succ(s start ) start 2 g = g = 7 6 g = 4 77

78 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s start is consistent with neighbors and eupanded. start 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. b. Update source nodes of changed edges.. Repeat from 2. 78

79 Handling Weight hanges Locally Self-consistent graph: g(s) = min s E Succ(s) (g(s) + c(s,s)) May no longer be true When edge Weights change 2 g = 2 g = g = 7 6 g = 4 g = 7 g = 4 79

80 Handling Weight hanges Locally hanges propagate to predecessors. For efficient search, update lo est cost nodes first. - Use a priority queue like A*. Update nodes until the goal is first eupanded. 2 g = g = 7 g = 4 80

81 Handling Weight hanges Locally Store an additional value: Local inconsistency: rhs(s) t g(s) rhs(s) = min (g(s) + c(s,s)) s E Succ(s) 2 g = g = 7 rhs = g = 4 8

82 Handling Weight hanges Locally Store an additional value: Local inconsistency rhs(s) t g(s) rhs(s) = min (g(s) + c(s,s)) s E Succ(s) 2 g = g = 7 rhs = g = 4 82

83 Local Inconsistencies Signal recomputation is necessary for node and predecessors. Locally overconsistent: g(s) > rhs(s) 2. Locally underconsistent: g(s) < rhs(s) 83

84 Updating and Eupanding odes Update by recomputing rhs and placing the node on the priority queue if locally inconsistent. s g = 7 rhs = g = g = 4 84

85 Updating and Eupanding odes Update by recomputing rhs and placing the node on the priority queue if locally inconsistent. s g = 7 rhs = 2 g = g = 4 85

86 Updating and Eupanding odes Update by recomputing rhs and placing the node on the priority queue if locally inconsistent. 2 g = =..., s,... s g = 7 rhs = g = 4 86

87 Updating and Eupanding odes Eupand by taking it off the priority queue and changing g. Eupand in order of total cost: ( min g(s), rhs(s) + h(s,s ), min g(s), rhs(s) ) start 2 g = s g = 7 rhs = g = 4 h = 4 87

88 Updating and Eupanding odes Eupand by taking it off the priority queue and changing g. Eupand in order of total cost: ( min g(s), rhs(s) + h(s,s ), min g(s), rhs(s) ) start 2 g = s g = rhs = g = 4 h = 4 88

89 89

90 g(s) > rhs(s) (Overconsistent): e path cost rhs(s) is better than the old path cost g(s). Immediately update g(s) to rhs(s) and propagate to all predeessors - Set g(s) = rhs(s) - Update all predecessors of s Verteu is no locally consistent and will remain that way. 90

91 g(s) < rhs(s) (Underconsistent): Old path cost g(s) is better than ne path cost rhs(s). Delay vertex expansion and propagate to all predeessors - Set g(s) = - Update all predecessors of s and s itself Verteu is no locally consistent or locally overconsistent. It Will remain on the queue until rhs(s) is the neut best cost. 9

92 g(s) < rhs(s) (Underconsistent): g = 9 s rhs = 9 2 h = 4 2 g = s 6 g = 7 rhs = 7 g = 4 h = 4

93 g(s) < rhs(s) (Underconsistent): Assume after update: s 2 g = 9 rhs = h = 4 g = s 6 g = 7 rhs = 0 h = 4 g = 4 93

94 g(s) < rhs(s) (Underconsistent): Assume after update: s 2 g = 9 rhs = h = 4 g = s 6 g = 7 rhs = 0 h = 4 g = 4 94

95 g(s) < rhs(s) (Underconsistent): s 2 g = 9 rhs = h = 4 g = s 6 g = 7 rhs = 0 h = 4 g = 4 95

96 g(s) < rhs(s) (Underconsistent): s 2 g = 9 rhs = h = 4 g = s g = rhs = 0 h = 4 6 g = 4 96

97 g(s) < rhs(s) (Underconsistent): s 2 g = 9 rhs = h = 4 g = s 6 g = 4 97

98 g(s) < rhs(s) (Underconsistent): s 2 g = rhs = h = 4 g = s g = rhs = 0 h = 4 6 g = 4 98

99 g(s) < rhs(s) (Underconsistent): s 2 g = rhs = h = 4 g = s g = rhs = 9 h = 4 6 g = 4 99

100 g(s) < rhs(s) (Underconsistent): s 2 g = rhs = h = 4 s g = g = rhs = 9 h = 4 6 g = 4 00

101 g(s) < rhs(s) (Underconsistent): s 2 g = rhs = h = 4 g = s g = 9 rhs = 9 h = 4 6 g = 4 0

102 g(s) < rhs(s) (Underconsistent): s 2 g = rhs = h = 4 g = s g = 9 rhs = 9 h = 4 6 g = 4 02

103 Overall D* Lite:. Initialize all nodes as uneupanded. 2. est-first search until s start is consistent With neighbors and eupanded. 3. Move to neut best verteu. 4. If any edge costs change: a. rack ho heuristics have changed. b. Update source nodes of changed edges.. Repeat from 2. 03

104 arrying Over the Priority Queue After a path is found, the priority queue is not empty. he value on the priority queue is: ( min g(s), rhs(s) + h(s,s ), min g(s), rhs(s) ) start hen s is different, the heuristics are different start 04

105 arrying Over the Priority Queue s s : All heuristics lo ered by at most h(s,s ). last start last start hen adding ne nodes to the queue, increase total cost by h(s last,s ). start Increase k = k + h(s last,s ): m m start ( min g(s), rhs(s) + h(s,s ) + k, min g(s), rhs(s) ) start m 05

106 Overall D* Lite:. Initialize all g(s) =, rhs(s t s goal ) =, rhs(s goal ) = 0, k = 0, s last = s. 2. est-first search until s is locally consistent and eupanded. start 3. Move so s = argmin (c(s,s) + g(s)). start s E Succ(s start ) start 4. If any edge costs change: a. k = k + h(s last,s ). m m start b. Update rhs and queue position for source nodes of changed edges.. Repeat from 2. m start Al ays sort by ( min g(s), rhs(s) + h(s,s ) + k, min g(s), rhs(s) ) start m 06

107 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 07

108 Overall D* Lite:. Initialize all g(s) =, rhs(s t s goal ) =, rhs(s goal ) = 0, k = 0, s last = s. 2. est-first search until s is locally consistent and eupanded. start 3. Move so s = argmin (c(s,s) + g(s)). start s E Succ(s start ) start 4. If any edge costs change: a. k = k + h(s last,s ). m m start b. Update rhs and queue position for source nodes of changed edges.. Repeat from 2. m start Al ays sort by ( min g(s), rhs(s) + h(s,s ) + k, min g(s), rhs(s) ) start m 08

109 D* Lite Euample Heuristic: h(a, ) = minimum number of nodes to get from A to o h(node, start) ritten in each node idirectional graph Edge eights labeled 0 3 Robot starts here Goal node 09

110 k = 0 m 0

111 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) 2 k = 0 m g rhs = 0 (deueued)

112 D* Lite Euample m k = 0 2. Plan initial path (A*) dequeue node In this euample, the presence of a key <u,y> indicates that the update g(node) node is in the min priority queue. update rhs(neighbor) 2 queue inconsistent neighbors g rhs = 0 (deueued)

113 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) update rhs(neighbor) queue inconsistent neighbors g rhs = (deueued) k = 0 m g = 0 rhs = 0 3 rhs = 0 <2,0>

114 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) update rhs(neighbor) queue inconsistent neighbors g rhs = (deueued) k = 0 m g = 0 rhs = 0 ote that this rhs decreased, and the nodes key changed accordingly.

115 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) update rhs(neighbor) queue inconsistent neighbors g = rhs = g 2 rhs = 2 (deueued) 2 0 k = 0 m g = 0 rhs = 0 5

116 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) update rhs(neighbor) queue inconsistent neighbors g = rhs = his nodes rhs as updated, but the value stayed the same. 0 3 g 2 rhs = 2 (deueued) his nodes rhs as updated, but the value stayed the same. k m = 0 g = 0 rhs = 0

117 D* Lite Euample 2. Plan initial path (A*) dequeue node update g(node) update rhs(neighbor) queue inconsistent neighbors g = rhs = 0 3 g 3 rhs = 3 (deueued) g = 2 rhs = k = 0 m g = 0 rhs = 0 7

118 D* Lite Euample Found shortest path to goal (edges in blue) g = rhs = 2 k = 0 m 0 3 g = 3 rhs = 3 g = 2 rhs = g = 0 rhs = 0 8

119 D* Lite Euample 3. Robot moves (green) update heuristics for ne start node maintain queue from before g = rhs = k = 0 m g = 0 rhs = 0 g = 3 rhs = 3 g = 2 rhs = 2 9

120 g = rhs = m g = 0 rhs = 0 g = 3 rhs = 3 g = 2 rhs = 2 20

121 m g = 3 rhs = g = 2 rhs 4 <3,2> 2 0 g = 0 rhs = 0 rhs(goal) = 0, al ays

122 D* Lite Euample. Repeat from 2 2. Replan path Underconsistent k m = g = 3 g = 2 rhs = 3 rhs = 4 Update rhs of g = 0 rhs = 0 <3,2> neighbors, but no values change. 22

123 D* Lite Euample 2. Replan path k = m g = 3 rhs = 3 Underconsistent g = 0 rhs = 0 23

124 D* Lite Euample 2. Replan path k = m g = 0 rhs = 0 24

125 D* Lite Euample 2. Replan path k = m g = 0 rhs = 0 25

126 D* Lite Euample 2. Replan path k = m g = 0 rhs = 0 26

127 D* Lite Euample 2. Replan path k = m g = 0 rhs = 0 27 g rhs = 0 (deueued)

128 D* Lite Euample 2. Replan path k = m g rhs = 0 (deueued) g = 0 rhs = 0

129 D* Lite Euample 2. Replan path k = m 0 2 g rhs = (deueued) 0 g = 0 rhs = 0 g = 0 rhs = 0 29

130 D* Lite Euample 2. Replan path k = m g = 0 rhs = 0 g rhs = (deueued) g = 0 rhs = 0 30

131 D* Lite Euample Found ne shortest path (blue) from current start (green) to goal (red) k = m 0 2 g = rhs = 0 g = 0 rhs = 0 g = 0 rhs = 0 3

132 D* Lite Euample 3. Robot moves (green) update heuristics for ne start node maintain queue from before k = m 2 g = rhs = 0 0 g = 0 rhs = 0 g = 0 rhs = 0 32

133 2 m 2 g = rhs = 0 0 g = 0 rhs = 0 g = 0 rhs = 0 33

134 2 m g = 0 rhs = 0 34 g = 0 rhs = 0

135 D* Lite Euample. Repeat from 2 2. Replan path g rhs = (deueued) k = 2 m g = 0 rhs = 0 g = 0 rhs = 0 35

136 D* Lite Euample 2. Replan path g = 0 rhs = 0 36 g = 0 rhs 2 <4,2>

137 D* Lite Euample 2. Replan path g = rhs = k = 2 m 2 0 g = 0 rhs = 0 g = rhs = g = rhs = <4,> 37 0 g 2 rhs = 2 (deueued)

138 D* Lite Euample Found ne shortest path g = rhs = k = 2 m 2 g = rhs = g = rhs = <4,> 38 0 g = 2 rhs = 2 0 g = 0 rhs = 0

139 D* Lite Euample 3. Robot moves update heuristics for ne start node maintain queue from before g = rhs = 0 k = 2 m 2 0 g = 0 rhs = 0 g = rhs = g = rhs = <4,> g = 2 rhs = 2 39

140 g = rhs = 0 3 m 2 g = 0 rhs = 0 g = rhs = g = rhs = <4,> g = 2 rhs = 2 40

141 g = rhs = 0 3 m 2 g = 0 rhs = 0 g = rhs 2 <7,2> g = rhs 2 <6,> 4 g = 2 rhs <6,2>

142 D* Lite Euample. Repeat from 2 2. Replan path o need to dequeue any nodes because the smallest key (<6,2>) is greater than key(start) (<4,>). 2 g = rhs = 2 <7,2> g = rhs = 2 <6,> 42 g = rhs = 0 g = 2 rhs = <6,2> k = 3 m g = 0 rhs = 0

143 D* Lite Euample 3. Robot moves Robot reaches goal: Done g = rhs = k = 3 m g = 0 rhs = 0 g = rhs = 2 <7,2> g = rhs = 2 <6,> 43 g = 2 rhs = <6,2>

144 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 44

145 Efficiency A* eupands each node at most once D* Lite (and LPA*) eupands each node at most t ice o ut ill in most cases eupand fe er nodes than A* A* might performs better if: o o hanges are close to start node Large changes to the graph 45

146 hen to use incremental path planning Goal 46

147 hen to use incremental path planning Goal 47

148 hen to use incremental path planning Goal 48

149 hen to use incremental path planning Goal 49

150 hen to use incremental path planning 50

151 hen to use incremental path planning Goal 5

152 hen to use incremental path planning Goal 52

153 hen to use incremental path planning Goal 53

154 hen to use incremental path planning Goal 54

155 hen to use incremental path planning 55

156 hen to use incremental path planning 56

157 hen to use incremental path planning 57

158 hen to use incremental path planning Goal 58

159 hen to use incremental path planning Goal 59

160 hen to use incremental path planning Goal 60

161 hen to use incremental path planning Goal 6

162 hen to use incremental path planning Goal 62

163 hen to use incremental path planning Goal 63

164 hen to use incremental path planning Goal 64

165 hen to use incremental path planning Goal 65

166 hen to use incremental path planning Goal 66

167 hen to use incremental path planning Goal 67

168 hen to use incremental path planning 68

169 hen to use incremental path planning 69

170 hen to use incremental path planning Goal 70

171 hen to use incremental path planning Goal 7

172 hen to use incremental path planning Goal 72

173 hen to use incremental path planning Goal 73

174 hen to use incremental path planning Goal 74

175 hen to use incremental path planning Goal 75

176 hen to use incremental path planning Goal 76

177 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 77

178 Greedy mapping 78

179 Greedy mapping Goal 79

180 Greedy mapping Goal 80

181 Greedy mapping Goal 8

182 Greedy mapping Goal 82

183 Greedy mapping Goal 83

184 Greedy mapping Goal 84

185 Greedy mapping Goal 85

186 Greedy mapping Goal 86

187 Greedy mapping Goal 87

188 Greedy mapping Goal 88

189 Greedy mapping Goal 89

190 Greedy mapping Goal 90

191 Greedy mapping Goal 9

192 Greedy mapping Goal 92

193 Greedy mapping Goal 93

194 Greedy mapping Goal 94

195 Greedy mapping Goal 95

196 Greedy mapping Goal 96

197 Anytime Dynamic A* (AD*) ombines the benefits of anytime and incremental planners In the real orld: o hanging environment (incremental planners) o Agents need to act upon decisions quickly (anytime planners) Ref: Likhachev, M., Ferguson, D. I., Gordon, G. J., Stentz, A., & hrun, S. (200, June). Anytime Dynamic A*: An Anytime, Replanning Algorithm. In ICAPS(pp ). 97

198 Anytime Planners Usually start off by computing a highly suboptimal solution o hen improves the solution over time A* ith inconsistent heuristics for euample o Inflated heuristic values give substantial speed-up o Inflated heuristics make the algorithm prefer to keep eupanding paths o Use incrementally less inflated heuristics to approach optimality 98

199 Anytime Dynamic A* (AD*) Starts off by setting a sufficiently high inflation factor o Generates suboptimal plan quickly Decreases inflation factor to approach optimality hen changes to edge costs are detected the current solution is repaired o If the changes are substantial the inflation factor is increased to generate a ne plan quickly 99

200 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 200

201 Application to mobile robotics This image in the public domain. 20 Reference: Lecture given by Michal p in 2.66 and paper currently in revie

202 Application to mobile robotics This image in the public domain. 202 Reference: Lecture given by Michal p in 2.66 and paper currently in revie

203 Holonomic System onholonomic System o differential constraints Differential constraints 203

204 Methods ell decomposition Visibility graph Sampling-based roadmap construction 204

205 ell Decomposition 205

206 ell Decomposition 206

207 ell Decomposition 207

208 ell Decomposition 208

209 ell Decomposition 209

210 ell Decomposition 20

211 ell Decomposition 2

212 ell Decomposition 22

213 ell Decomposition 23

214 ell Decomposition 24

215 ell Decomposition 25

216 ell Decomposition 26

217 ell Decomposition 27

218 ell Decomposition 28

219 ell Decomposition 29

220 Visibility Graph 220

221 Visibility Graph 22

222 Visibility Graph 222

223 Visibility Graph 223

224 Visibility Graph 224

225 Visibility Graph 225

226 Sampling-based Roadmap onstruction Deterministic sampling Random sampling 226

227 Sampling-based Roadmap onstruction 227

228 onnecting Samples Steering function used o Steer(a, b) gives feasible path bet een samples o onholonomic Dubins path o Shortest path bet een t o points o an be constructed of mauimum curvature and straight line segments RSR, RSL, LSR, LSL, RLR or LRL 228

229 Outline Motivation Incremental Search The D* Lite Algorithm D* Lite Example When to Use Incremental Path Planning? Algorithm Extensions and Related Topics Application to Mobile Robotics 229

230 230

231 ackup 23

232 Incremental Path Planning Dynamic A* (D*) - Stentz, 994 o Initial combination on A* and incremental search nor mobile robot path planning. DynamicsSWSF-FP - Ramalingam and Reps, 99 o Incremental search technique. Linelong Planning A* (LPA*) - Koenig and Likhachev, 200 o Generalizes A* and DynamicsSWSF-FP nor indeninite re-planning on ninite graph. D* Lite - Koenig and Likhachev, 2002 o o Extends LPA* to provide D* nunctionality nor mobile robots. Simpler normulation than D*. More recent extensions to D* and D* Lite, including Anytime D* (AD*), Field D*, etc. Other methods

233 D* Algorithm 233

234 LPA* Algorithm From Koenig, Likhachev, & Furcy (2004) Courtesy of Elsevier, Inc., Used with permission. Source: Figures 3 and 4 in Keonig, Sven, M. Likhachev, and D. Furcy. "Lifelong Planning A*." In Artificial Intelligence, 55 (- 2):

235 Revie of A* Use admissible heuristic, h(s) < h*(s), to guide search. Keep track of total cost distance from start, g(s). Order node eupansions using priority ueue, with priorities (s) g(s) h(s). Avoid re-eupanding nodes by using eupanded list. etter heuristics (ho closely h(s) approuimates of h*(s)) improve search speed. Guaranteed to return optimal solution if one euists. 235

236 RHS Values One-step look-ahead on g-values, rhs(s) = 0 if s is beginning node of search, other ise: rhs(s) = min (g(s') + c(s', s)) s' ( pred(s) Potentially better innormed than g-value anter changes to search graph. Note: term comes nrom grammar rules used in DynamicsSWSF-FP algorithm, no other signinicance. 236

237 Local onsistency Tells us which nodes may need g-values updated in order to nind shortest path. Node s is locally consistent inn: g(s) = rhs(s) Node s is locally overconsistent inn: g(s) > rhs(s) Node s is locally underconsistent inn: g(s) < rhs(s) Initially, all nodes are locally consistent with g(s) = rhs(s) =, with exception on start node, rhs(s ) = 0 and g(s ) = start start 237

238 238

239 Anytime Dynamic A* (AD*) American Association for Artificial Intelligence. All rights reserved. This content is excluded from our Creative Commons license. For more information, see Fig: Likhachev, M., Ferguson, D. I., Gordon, G. J., Stentz, A., & hrun, S. (200, June). Anytime Dynamic A*: An Anytime, Replanning Algorithm. In ICAPS(pp ). 239

240 MIT OpenCourseWare J / 6.834J Cognitive Robotics Spring 206 For information about citing these materials or our Terms of Use, visit:

Incremental A. S. Koenig and M. Likhachev Georgia Institute of Technology College of Computing Atlanta, GA skoenig,

Incremental A. S. Koenig and M. Likhachev Georgia Institute of Technology College of Computing Atlanta, GA skoenig, Incremental A S. Koenig and M. Likhachev Georgia Institute of Technology College of Computing Atlanta, GA 30312-0280 skoenig, mlikhach @cc.gatech.edu Abstract Incremental search techniques find optimal

More information

Anytime Search in Dynamic Graphs

Anytime Search in Dynamic Graphs Anytime Search in Dynamic Graphs Maxim Likhachev a, Dave Ferguson b,c, Geoff Gordon c, Anthony Stentz c, and Sebastian Thrun d a Computer and Information Science, University of Pennsylvania, Philadelphia,

More information

Anytime search in dynamic graphs

Anytime search in dynamic graphs University of Pennsylvania ScholarlyCommons Lab Papers (GRASP) General Robotics, Automation, Sensing and Perception Laboratory 9-2008 Anytime search in dynamic graphs Maxim Likhachev University of Pennsylvania,

More information

3D Field D*: Improved Path Planning and Replanning in Three Dimensions

3D Field D*: Improved Path Planning and Replanning in Three Dimensions Proceedings of the 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems October 9-15, 2006, Beijing, China 3D Field D*: Improved Path Planning and Replanning in Three Dimensions Joseph

More information

Search-based Planning with Motion Primitives. Maxim Likhachev Carnegie Mellon University

Search-based Planning with Motion Primitives. Maxim Likhachev Carnegie Mellon University Search-based Planning with Motion Primitives Maxim Likhachev Carnegie Mellon University generate a graph representation of the planning problem search the graph for a solution What is Search-based Planning

More information

Autonomous Navigation in Unknown Environments via Language Grounding

Autonomous Navigation in Unknown Environments via Language Grounding Autonomous Navigation in Unknown Environments via Language Grounding Koushik (kbhavani) Aditya (avmandal) Sanjay (svnaraya) Mentor Jean Oh Introduction As robots become an integral part of various domains

More information

D* Lite. Sven Koenig College of Computing Georgia Institute of Technology Atlanta, GA

D* Lite. Sven Koenig College of Computing Georgia Institute of Technology Atlanta, GA From: I-2 Proceedings. Copyright 22, I (www.aaai.org). ll rights reserved. D* Lite Sven Koenig College of Computing Georgia Institute of Technology tlanta, G -2 skoenig@cc.gatech.edu Maxim Likhachev School

More information

Reusing Previously Found A* Paths for Fast Goal-Directed Navigation in Dynamic Terrain

Reusing Previously Found A* Paths for Fast Goal-Directed Navigation in Dynamic Terrain Reusing Previously Found A* Paths for Fast Goal-Directed Navigation in Dynamic Terrain Carlos Hernández Depto. de Ingeniería Informática Univ. Católica de la Ssma. Concepción Concepción, Chile Roberto

More information

Anytime Path Planning and Replanning in Dynamic Environments

Anytime Path Planning and Replanning in Dynamic Environments Anytime Path Planning and Replanning in Dynamic Environments Jur van den Berg Department of Information and Computing Sciences Universiteit Utrecht The Netherlands berg@cs.uu.nl Dave Ferguson and James

More information

ANA*: Anytime Nonparametric A*

ANA*: Anytime Nonparametric A* ANA*: Anytime Nonparametric A* Jur van den Berg 1 and Rajat Shah 2 and Arthur Huang 2 and Ken Goldberg 2 1 University of North Carolina at Chapel Hill. E-mail: berg@cs.unc.edu. 2 University of California,

More information

Search-Based Planning with Provable Suboptimality Bounds for Continuous State Spaces

Search-Based Planning with Provable Suboptimality Bounds for Continuous State Spaces Search-Based Planning with Provable Suboptimality Bounds Juan Pablo Gonzalez for Continuous State Spaces Maxim Likhachev Autonomous Perception Research Robotics Institute General Dynamics Robotic Systems

More information

Computer Game Programming Basic Path Finding

Computer Game Programming Basic Path Finding 15-466 Computer Game Programming Basic Path Finding Robotics Institute Path Planning Sven Path Planning needs to be very fast (especially for games with many characters) needs to generate believable paths

More information

Real-Time Adaptive A*

Real-Time Adaptive A* Real-Time Adaptive A* Sven Koenig Computer Science Department University of Southern California Los Angeles, CA - skoenig@usc.edu Maxim Likhachev Computer Science Department Carnegie Mellon University

More information

Improving MPGAA* for Extended Visibility Ranges

Improving MPGAA* for Extended Visibility Ranges Proceedings of the Twenty-Seventh International Conference on Automated Planning and Scheduling (ICAPS 2017) Improving MPGAA* for Extended Visibility Ranges Carlos Hernández Depto. de Ciencias de la Ingeniería

More information

Introduction to Multi-Agent Programming

Introduction to Multi-Agent Programming Introduction to Multi-Agent Programming 4. Search algorithms and Pathfinding Uninformed & informed search, online search, ResQ Freiburg path planner Alexander Kleiner, Bernhard Nebel Contents Problem-Solving

More information

1 Project: Fast Trajectory Replanning for Computer Games

1 Project: Fast Trajectory Replanning for Computer Games Project: Fast rajectory Replanning for omputer Games Figure : otal nnihilation by avedog onsider characters in real-time computer games, such as otal nnihilation shown in Figure. o make them easy to control,

More information

D* lite algorithm based path planning of mobile robot in static Environment

D* lite algorithm based path planning of mobile robot in static Environment D* lite algorithm based path planning of mobile robot in static Environment Pradipta kumar Das 1, Dr. S.N. Patro 2,C.N. Panda 3, Bunil Balabantaray 4 Dhaneswar Rath Institute of Engineering and Management

More information

Cognitive Robotics Robot Motion Planning Matteo Matteucci

Cognitive Robotics Robot Motion Planning Matteo Matteucci Cognitive Robotics Robot Motion Planning Robot Motion Planning eminently necessary since, by definition, a robot accomplishes tasks by moving in the real world. J.-C. Latombe (1991) Robot Motion Planning

More information

Visual Navigation for Flying Robots. Motion Planning

Visual Navigation for Flying Robots. Motion Planning Computer Vision Group Prof. Daniel Cremers Visual Navigation for Flying Robots Motion Planning Dr. Jürgen Sturm Motivation: Flying Through Forests 3 1 2 Visual Navigation for Flying Robots 2 Motion Planning

More information

Robot Motion Control Matteo Matteucci

Robot Motion Control Matteo Matteucci Robot Motion Control Open loop control A mobile robot is meant to move from one place to another Pre-compute a smooth trajectory based on motion segments (e.g., line and circle segments) from start to

More information

ANA*: Anytime Nonparametric A*

ANA*: Anytime Nonparametric A* ANA*: Anytime Nonparametric A* Jur van den Berg 1 and Rajat Shah 2 and Arthur Huang 2 and Ken Goldberg 2 1 University of North Carolina at Chapel Hill. E-mail: berg@cs.unc.edu. 2 University of California,

More information

A Survey and Classification of A* based Best-First Heuristic Search Algorithms

A Survey and Classification of A* based Best-First Heuristic Search Algorithms A Survey and Classification of A* based Best-First Heuristic Search Algorithms Luis Henrique Oliveira Rios and Luiz Chaimowicz Departamento de Ciência da Computação Universidade Federal de Minas Gerais

More information

CS 387/680: GAME AI PATHFINDING

CS 387/680: GAME AI PATHFINDING CS 387/680: GAME AI PATHFINDING 4/16/2015 Instructor: Santiago Ontañón santi@cs.drexel.edu Class website: https://www.cs.drexel.edu/~santi/teaching/2015/cs387/intro.html Reminders Check BBVista site for

More information

Steve Levine. Programs with Flexible Time. When? Tuesday, Feb 16 th. courtesy of JPL

Steve Levine. Programs with Flexible Time. When? Tuesday, Feb 16 th. courtesy of JPL Programs with Flexible Time When? Contributions: Brian Williams Patrick Conrad Simon Fang Paul Morris Nicola Muscettola Pedro Santana Julie Shah John Stedl Andrew Wang Steve Levine Tuesday, Feb 16 th courtesy

More information

Sub-Optimal Heuristic Search ARA* and Experience Graphs

Sub-Optimal Heuristic Search ARA* and Experience Graphs Robot Autonomy (16-662, S13) Lecture#09 (Wednesday Feb 13, 2013) Lecture by: Micheal Phillips Sub-Optimal Heuristic Search ARA* and Experience Graphs Scribes: S.Selvam Raju and Adam Lebovitz Contents 1

More information

CSE-571. Deterministic Path Planning in Robotics. Maxim Likhachev 1. Motion/Path Planning. Courtesy of Maxim Likhachev University of Pennsylvania

CSE-571. Deterministic Path Planning in Robotics. Maxim Likhachev 1. Motion/Path Planning. Courtesy of Maxim Likhachev University of Pennsylvania Motion/Path Planning CSE-57 Deterministic Path Planning in Robotics Courtesy of Maxim Likhachev University of Pennsylvania Tas k : find a feasible (and cost-minimal) path/motion from the current configuration

More information

Real-Time Search in Dynamic Worlds

Real-Time Search in Dynamic Worlds Real-Time Search in Dynamic Worlds David M. Bond, Niels A. Widger, Wheeler Ruml Department of Computer Science University of New Hampshire Durham, NH 0384 USA david.bond at iol.unh.edu niels.widger, ruml

More information

Fast Replanning for Navigation in Unknown Terrain

Fast Replanning for Navigation in Unknown Terrain I TRNSTIONS ON ROOTIS N UTOMTION, VOL. XX, NO. Y, MONTH 00 Fast Replanning for Navigation in Unknown Terrain Sven Koenig, Member, I, Maxim Likhachev bstract Mobile robots often operate in domains that

More information

Planning for Markov Decision Processes with Sparse Stochasticity

Planning for Markov Decision Processes with Sparse Stochasticity Planning for Markov Decision Processes with Sparse Stochasticity Maxim Likhachev Geoff Gordon Sebastian Thrun School of Computer Science School of Computer Science Dept. of Computer Science Carnegie Mellon

More information

R* Search. Anthony Stentz The Robotics Institute Carnegie Mellon University Pittsburgh, PA

R* Search. Anthony Stentz The Robotics Institute Carnegie Mellon University Pittsburgh, PA R* Search Maxim Likhachev Computer and Information Science University of Pennsylvania Philadelphia, PA 19104 maximl@seas.upenn.edu Anthony Stentz The Robotics Institute Carnegie Mellon University Pittsburgh,

More information

MOBILE robots often operate in domains that are only incompletely

MOBILE robots often operate in domains that are only incompletely 354 IEEE TRANSACTIONS ON ROBOTICS, VOL. 21, NO. 3, JUNE 2005 Fast Replanning for Navigation in Unknown Terrain Sven Koenig, Senior Member, IEEE, and Maxim Likhachev Abstract Mobile robots often operate

More information

Anytime Tree-Restoring Weighted A* Graph Search

Anytime Tree-Restoring Weighted A* Graph Search Proceedings of the Seventh Annual Symposium on Combinatorial Search (SoCS 2014) Anytime Tree-Restoring Weighted A* Graph Search Kalin Gochev University of Pennsylvania Alla Safonova University of Pennsylvania

More information

Any-Angle Search Case Study: Theta* slides by Alex Nash / with contributions by Sven Koenig

Any-Angle Search Case Study: Theta* slides by Alex Nash / with contributions by Sven Koenig Any-Angle Search Case Study: Theta* slides by Alex Nash anash@usc.edu / alexwnash@gmail.com with contributions by Sven Koenig skoenig@usc.edu Table of Contents o Introduction o Analysis of Path Lengths

More information

Path Planning. Marcello Restelli. Dipartimento di Elettronica e Informazione Politecnico di Milano tel:

Path Planning. Marcello Restelli. Dipartimento di Elettronica e Informazione Politecnico di Milano   tel: Marcello Restelli Dipartimento di Elettronica e Informazione Politecnico di Milano email: restelli@elet.polimi.it tel: 02 2399 3470 Path Planning Robotica for Computer Engineering students A.A. 2006/2007

More information

Single-Robot and Multi-Robot Path Planning with Quality Guarantees Cognitive Robotics Summer School 2017

Single-Robot and Multi-Robot Path Planning with Quality Guarantees Cognitive Robotics Summer School 2017 Single-Robot and Multi-Robot Path Planning with Quality Guarantees Cognitive Robotics Summer School 207 Sven Koenig, the awesome people in his research group and their awesome collaborators University

More information

Geometric Path Planning McGill COMP 765 Oct 12 th, 2017

Geometric Path Planning McGill COMP 765 Oct 12 th, 2017 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

More information

Sub-optimal Heuristic Search and its Application to Planning for Manipulation

Sub-optimal Heuristic Search and its Application to Planning for Manipulation Sub-optimal Heuristic Search and its Application to Planning for Manipulation Mike Phillips Slides adapted from Planning for Mobile Manipulation What planning tasks are there? robotic bartender demo at

More information

Autonomous robot motion path planning using shortest path planning algorithms

Autonomous robot motion path planning using shortest path planning algorithms IOSR Journal of Engineering (IOSRJEN) e-issn: 2250-3021, p-issn: 2278-8719 Vol. 3, Issue 1 (Jan. 2013), V1 PP 65-69 Autonomous robot motion path planning using shortest path planning algorithms T. Ravi

More information

Class Overview. Introduction to Artificial Intelligence COMP 3501 / COMP Lecture 2. Problem Solving Agents. Problem Solving Agents: Assumptions

Class Overview. Introduction to Artificial Intelligence COMP 3501 / COMP Lecture 2. Problem Solving Agents. Problem Solving Agents: Assumptions Class Overview COMP 3501 / COMP 4704-4 Lecture 2 Prof. JGH 318 Problem Solving Agents Problem Solving Agents: Assumptions Requires a goal Assume world is: Requires actions Observable What actions? Discrete

More information

6.141: Robotics systems and science Lecture 10: Motion Planning III

6.141: Robotics systems and science Lecture 10: Motion Planning III 6.141: Robotics systems and science Lecture 10: Motion Planning III Lecture Notes Prepared by N. Roy and D. Rus EECS/MIT Spring 2012 Reading: Chapter 3, and Craig: Robotics http://courses.csail.mit.edu/6.141/!

More information

Dynamic Fringe-Saving A*

Dynamic Fringe-Saving A* ynamic ringe-saving * Xiaoxun Sun William Yeoh Sven Koenig omputer Science epartment University of Southern alifornia Los ngeles, 90089-0781, US {xiaoxuns, wyeoh, skoenig}@usc.edu bstract ringe-saving

More information

Planning & Decision-making in Robotics Case Study: Planning for Autonomous Driving

Planning & Decision-making in Robotics Case Study: Planning for Autonomous Driving 16-782 Planning & Decision-making in Robotics Case Study: Planning for Autonomous Driving Maxim Likhachev Robotics Institute Carnegie Mellon University Typical Planning Architecture for Autonomous Vehicle

More information

6.141: Robotics systems and science Lecture 10: Implementing Motion Planning

6.141: Robotics systems and science Lecture 10: Implementing Motion Planning 6.141: Robotics systems and science Lecture 10: Implementing Motion Planning Lecture Notes Prepared by N. Roy and D. Rus EECS/MIT Spring 2011 Reading: Chapter 3, and Craig: Robotics http://courses.csail.mit.edu/6.141/!

More information

Solving Shortest Path Problems with Curvature Constraints Using Beamlets

Solving Shortest Path Problems with Curvature Constraints Using Beamlets Solving Shortest Path Problems with Curvature Constraints Using Beamlets Oktay Arslan, Panagiotis Tsiotras and Xiaoming Huo Abstract A new graph representation of a 2D environment is proposed in order

More information

Robotic Motion Planning: A* and D* Search

Robotic Motion Planning: A* and D* Search Robotic Motion Planning: A* and D* Search Robotics Institute 6-75 http://voronoi.sbp.ri.cmu.edu/~motion Howie Choset http://voronoi.sbp.ri.cmu.edu/~choset 6-75, Howie Choset with slides from G. Ayorker

More information

A Project on Fast Trajectory Replanning for Computer Games for Introduction to Artificial Intelligence Classes

A Project on Fast Trajectory Replanning for Computer Games for Introduction to Artificial Intelligence Classes Project on Fast rajectory Replanning for omputer Games for Introduction to rtificial Intelligence lasses Sven Koenig and William Yeoh epartment of omputer Science University of Southern alifornia Henry

More information

Multi-Scale LPA* with Low Worst-Case Complexity Guarantees

Multi-Scale LPA* with Low Worst-Case Complexity Guarantees Multi-Scale LPA* with Low Worst-Case Complexity Guarantees Yibiao Lu, Xiaoming Huo, Oktay Arslan, and Panagiotis Tsiotras Abstract In this paper we consider dynamic shortest pathplanning problems on a

More information

Approximate path planning. Computational Geometry csci3250 Laura Toma Bowdoin College

Approximate path planning. Computational Geometry csci3250 Laura Toma Bowdoin College Approximate path planning Computational Geometry csci3250 Laura Toma Bowdoin College Outline Path planning Combinatorial Approximate Combinatorial path planning Idea: Compute free C-space combinatorially

More information

Introduction to State-of-the-art Motion Planning Algorithms. Presented by Konstantinos Tsianos

Introduction to State-of-the-art Motion Planning Algorithms. Presented by Konstantinos Tsianos Introduction to State-of-the-art Motion Planning Algorithms Presented by Konstantinos Tsianos Robots need to move! Motion Robot motion must be continuous Geometric constraints Dynamic constraints Safety

More information

Dynamic Fringe-Saving A*

Dynamic Fringe-Saving A* ynamic ringe-saving * Xiaoxun Sun William Yeoh Sven Koenig omputer Science epartment University of Southern alifornia Los ngeles, 90089-0781, US {xiaoxuns, wyeoh, skoenig}@usc.edu STRT ringe-saving * is

More information

Incremental ARA*: An Incremental Anytime Search Algorithm for Moving-Target Search

Incremental ARA*: An Incremental Anytime Search Algorithm for Moving-Target Search Incremental R*: n Incremental nytime Search lgorithm for Moving-Target Search Xiaoxun Sun Tansel Uras Sven Koenig omputer Science epartment University of Southern alifornia 941 W 37th Street, Los ngeles,

More information

A Survey of Suboptimal Search Algorithms

A Survey of Suboptimal Search Algorithms A Survey of Search Algorithms Jordan T. Thayer and Wheeler Ruml jtd7, ruml at cs.unh.edu slides at: http://www.cs.unh.edu/ jtd7/papers/ Jordan Thayer and Wheeler Ruml (UNH) Search 1 / 28 This is Search

More information

A Comparison of Fast Search Methods for Real-Time Situated Agents

A Comparison of Fast Search Methods for Real-Time Situated Agents A Comparison of Fast Search Methods for Real-Time Situated Agents Sven Koenig Computer Science Department, University of Southern California 941 W 37th Street, Los Angeles, California 90089-0781 skoenig@usc.edu

More information

EE631 Cooperating Autonomous Mobile Robots

EE631 Cooperating Autonomous Mobile Robots EE631 Cooperating Autonomous Mobile Robots Lecture 3: Path Planning Algorithm Prof. Yi Guo ECE Dept. Plan Representing the Space Path Planning Methods A* Search Algorithm D* Search Algorithm Representing

More information

CS 5522: Artificial Intelligence II

CS 5522: Artificial Intelligence II CS 5522: Artificial Intelligence II Search Algorithms Instructor: Wei Xu Ohio State University [These slides were adapted from CS188 Intro to AI at UC Berkeley.] Today Agents that Plan Ahead Search Problems

More information

Planning, Execution and Learning Heuristics and Multi-Heuristic A*

Planning, Execution and Learning Heuristics and Multi-Heuristic A* 15-887 Planning, Execution and Learning Heuristics and Multi-Heuristic A* Maxim Likhachev Robotics Institute Carnegie Mellon University Design of Informative Heuristics Example problem: move picture frame

More information

Chapter 10 Motion Planning

Chapter 10 Motion Planning Chapter 10 Motion Planning Part 3 10.3 Real Time Global Motion Planning 1 Outline 10.3 Real Time Global Motion Planning 10.3.1 Introduction 10.3.2 Depth Limited Approaches 10.3.3 Anytime Approaches 10.3.4

More information

Jane Li. Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute

Jane Li. Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute Jane Li Assistant Professor Mechanical Engineering Department, Robotic Engineering Program Worcester Polytechnic Institute A search-algorithm prioritizes and expands the nodes in its open list items by

More information

Generalized Adaptive A*

Generalized Adaptive A* Generalized daptive * Xiaoxun Sun US omputer Science Los ngeles, alifornia xiaoxuns@usc.edu Sven Koenig US omputer Science Los ngeles, alifornia skoenig@usc.edu William Yeoh US omputer Science Los ngeles,

More information

Humanoid Robotics. Path Planning and Walking. Maren Bennewitz

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

More information

Theta*: Any-Angle Path Planning on Grids

Theta*: Any-Angle Path Planning on Grids Theta*: ny-ngle Path Planning on Grids Kenny Daniel lex Nash Sven Koenig omputer Science Department University of Southern alifornia Los ngeles, alifornia 90089-0781, US riel Felner Department of Information

More information

Motion Planning, Part IV Graph Search Part II. Howie Choset

Motion Planning, Part IV Graph Search Part II. Howie Choset Motion Planning, Part IV Graph Search Part II Howie Choset Map-Based Approaches: Properties of a roadmap: Roadmap Theory Accessibility: there exists a collision-free path from the start to the road map

More information

Speeding Up the Calculation of Heuristics for Heuristic Search-Based Planning

Speeding Up the Calculation of Heuristics for Heuristic Search-Based Planning From: AAAI-2 Proceedings. Copyright 2, AAAI (www.aaai.org). All rights reserved. Speeding Up the Calculation of Heuristics for Heuristic Search-Based Planning Yaxin Liu, Sven Koenig and David Furcy College

More information

ME/CS 132: Advanced Robotics: Navigation and Vision

ME/CS 132: Advanced Robotics: Navigation and Vision ME/CS 132: Advanced Robotics: Navigation and Vision Lecture #5: Search Algorithm 1 Yoshiaki Kuwata 4/12/2011 Lecture Overview Introduction Label Correcting Algorithm Core idea Depth-first search Breadth-first

More information

Beamlet-like Data Processing for Accelerated Path-Planning Using Multiscale Information of the Environment

Beamlet-like Data Processing for Accelerated Path-Planning Using Multiscale Information of the Environment Beamlet-like Data Processing for Accelerated Path-Planning Using Multiscale Information of the Environment Yibiao Lu, Xiaoming Huo, and Panagiotis Tsiotras Abstract We consider the deterministic path-planning

More information

Prediction-Based Path Planning with Obstacle Avoidance in Dynamic Target Environment

Prediction-Based Path Planning with Obstacle Avoidance in Dynamic Target Environment 48 Prediction-Based Path Planning with Obstacle Avoidance in Dynamic Target Environment Zahraa Y. Ibrahim Electrical Engineering Department University of Basrah Basrah, Iraq Abdulmuttalib T. Rashid Electrical

More information

Introduction to Information Science and Technology (IST) Part IV: Intelligent Machines and Robotics Planning

Introduction to Information Science and Technology (IST) Part IV: Intelligent Machines and Robotics Planning Introduction to Information Science and Technology (IST) Part IV: Intelligent Machines and Robotics Planning Sören Schwertfeger / 师泽仁 ShanghaiTech University ShanghaiTech University - SIST - 10.05.2017

More information

Efficient Interpolated Path Planning of Mobile Robots based on Occupancy Grid Maps

Efficient Interpolated Path Planning of Mobile Robots based on Occupancy Grid Maps Efficient Interpolated Path Planning of Mobile Robots based on Occupancy Grid Maps Marija Ðakulović Mijo Čikeš Ivan Petrović University of Zagreb, Faculty of Electrical Engineering and Computing, Department

More information

CS 460/560 Introduction to Computational Robotics Fall 2017, Rutgers University. Lecture 08 Extras. A* In More Detail. Instructor: Jingjin Yu

CS 460/560 Introduction to Computational Robotics Fall 2017, Rutgers University. Lecture 08 Extras. A* In More Detail. Instructor: Jingjin Yu CS 460/560 Introduction to Computational Robotics Fall 2017, Rutgers University Lecture 08 Extras A* In More Detail Instructor: Jingjin Yu Outline A* in more detail Admissible and consistent heuristic

More information

Motion Planning for a Point Robot (2/2) Point Robot on a Grid. Planning requires models. Point Robot on a Grid 1/18/2012.

Motion Planning for a Point Robot (2/2) Point Robot on a Grid. Planning requires models. Point Robot on a Grid 1/18/2012. Motion Planning for a Point Robot (/) Class scribing Position paper 1 Planning requires models Point Robot on a Grid The Bug algorithms are reactive motion strategies ; they are not motion planners To

More information

Advanced Robotics Path Planning & Navigation

Advanced Robotics Path Planning & Navigation Advanced Robotics Path Planning & Navigation 1 Agenda Motivation Basic Definitions Configuration Space Global Planning Local Planning Obstacle Avoidance ROS Navigation Stack 2 Literature Choset, Lynch,

More information

Sung-Eui Yoon ( 윤성의 )

Sung-Eui Yoon ( 윤성의 ) Path Planning for Point Robots Sung-Eui Yoon ( 윤성의 ) Course URL: http://sglab.kaist.ac.kr/~sungeui/mpa Class Objectives Motion planning framework Classic motion planning approaches 2 3 Configuration Space:

More information

Heuristic Search-Based Replanning

Heuristic Search-Based Replanning Heuristic Search-ased Replanning Sven Koenig avid Furcy olin auer ollege of omputing Georgia Institute of Technology tlanta, G 1-8 {skoenig, dfurcy, colinos}@cc.gatech.edu bstract Many real-world planning

More information

Basic Motion Planning Algorithms

Basic Motion Planning Algorithms Basic Motion Planning Algorithms Sohaib Younis Intelligent Robotics 7 January, 2013 1 Outline Introduction Environment Map Dijkstra's algorithm A* algorithm Summary 2 Introduction Definition: Motion Planning

More information

Robotics Tasks. CS 188: Artificial Intelligence Spring Manipulator Robots. Mobile Robots. Degrees of Freedom. Sensors and Effectors

Robotics Tasks. CS 188: Artificial Intelligence Spring Manipulator Robots. Mobile Robots. Degrees of Freedom. Sensors and Effectors CS 188: Artificial Intelligence Spring 2006 Lecture 5: Robot Motion Planning 1/31/2006 Dan Klein UC Berkeley Many slides from either Stuart Russell or Andrew Moore Motion planning (today) How to move from

More information

Informed Search A* Algorithm

Informed Search A* Algorithm Informed Search A* Algorithm CE417: Introduction to Artificial Intelligence Sharif University of Technology Spring 2018 Soleymani Artificial Intelligence: A Modern Approach, Chapter 3 Most slides have

More information

Chapter 10 Motion Planning

Chapter 10 Motion Planning Chapter 10 Motion Planning 10.1 Introduction Part 1 1 10.1 Introduction Outline 10.1.1 Introducing Motion Planning 10.1.2 Formulation of Motion Planning 10.1.3 Obstacle Free Motion Planning Summary 2 Hierarchy

More information

Robot Motion Planning

Robot Motion Planning Robot Motion Planning slides by Jan Faigl Department of Computer Science and Engineering Faculty of Electrical Engineering, Czech Technical University in Prague lecture A4M36PAH - Planning and Games Dpt.

More information

Speeding up Moving-Target Search

Speeding up Moving-Target Search Speeding up Moving-Target Search Sven Koenig US S epartment skoenig@usc.edu Maxim Likhachev MU Robotics nstitute maxim+@cs.cmu.edu Xiaoxun Sun US S epartment xiaoxuns@usc.edu STRT n this paper, we study

More information

Heuristic Search: A* CPSC 322 Search 4 January 19, Textbook 3.6 Taught by: Vasanth

Heuristic Search: A* CPSC 322 Search 4 January 19, Textbook 3.6 Taught by: Vasanth Heuristic Search: A* CPSC 322 Search 4 January 19, 2011 Textbook 3.6 Taught by: Vasanth 1 Lecture Overview Recap Search heuristics: admissibility and examples Recap of BestFS Heuristic search: A* 2 Example

More information

CMU-Q Lecture 4: Path Planning. Teacher: Gianni A. Di Caro

CMU-Q Lecture 4: Path Planning. Teacher: Gianni A. Di Caro CMU-Q 15-381 Lecture 4: Path Planning Teacher: Gianni A. Di Caro APPLICATION: MOTION PLANNING Path planning: computing a continuous sequence ( a path ) of configurations (states) between an initial configuration

More information

Search and Games. Adi Botea. ANU Summer Schools in Logic and Learning February, 2009

Search and Games. Adi Botea. ANU Summer Schools in Logic and Learning February, 2009 Search and Games Adi Botea ANU Summer Schools in Logic and Learning February, 2009 Outline 1 Introduction 2 Problem Representation 3 Uninformed Search 4 Informed Search 5 Hierarchical Abstraction Outline

More information

Planning Techniques for Robotics Planning Representations: Skeletonization- and Grid-based Graphs

Planning Techniques for Robotics Planning Representations: Skeletonization- and Grid-based Graphs 16-350 Planning Techniques for Robotics Planning Representations: Skeletonization- and Grid-based Graphs Maxim Likhachev Robotics Institute 2D Planning for Omnidirectional Point Robot Planning for omnidirectional

More information

Notes. Video Game AI: Lecture 5 Planning for Pathfinding. Lecture Overview. Knowledge vs Search. Jonathan Schaeffer this Friday

Notes. Video Game AI: Lecture 5 Planning for Pathfinding. Lecture Overview. Knowledge vs Search. Jonathan Schaeffer this Friday Notes Video Game AI: Lecture 5 Planning for Pathfinding Nathan Sturtevant COMP 3705 Jonathan Schaeffer this Friday Planning vs localization We cover planning today Localization is just mapping a real-valued

More information

CS 387/680: GAME AI PATHFINDING

CS 387/680: GAME AI PATHFINDING CS 387/680: GAME AI PATHFINDING 4/14/2014 Instructor: Santiago Ontañón santi@cs.drexel.edu TA: Alberto Uriarte office hours: Tuesday 4-6pm, Cyber Learning Center Class website: https://www.cs.drexel.edu/~santi/teaching/2014/cs387-680/intro.html

More information

CS 343H: Artificial Intelligence

CS 343H: Artificial Intelligence CS 343H: Artificial Intelligence Lecture 4: Informed Search 1/23/2014 Slides courtesy of Dan Klein at UC-Berkeley Unless otherwise noted Today Informed search Heuristics Greedy search A* search Graph search

More information

Planning in Domains with Cost Function Dependent Actions

Planning in Domains with Cost Function Dependent Actions Planning in Domains with Cost Function Dependent Actions Mike Phillips and Maxim Likhachev mlphilli@andrew.cmu.edu, maxim@cs.cmu.edu Carnegie Mellon University Content Area Constraints, Satisfiability,

More information

Theta*: Any-Angle Path Planning on Grids

Theta*: Any-Angle Path Planning on Grids Theta*: ny-ngle Path Planning on Grids lex Nash and Kenny Daniel and Sven Koenig riel Felner omputer Science Department Department of Information Systems Engineering University of Southern alifornia en-gurion

More information

On the modeling and the optimal motion planning of manipulators via a modified D* Lite search algorithm

On the modeling and the optimal motion planning of manipulators via a modified D* Lite search algorithm On the modeling and the optimal motion planning of manipulators via a modified D* Lite search algorithm AMIR FEIZOLLAHI and RENE V. MAYORGA Department of Industrial Systems Engineering University of Regina

More information

Bounded Suboptimal Search: A Direct Approach Using Inadmissible Estimates

Bounded Suboptimal Search: A Direct Approach Using Inadmissible Estimates Bounded Suboptimal Search: A Direct Approach Using Inadmissible Estimates Jordan T. Thayer and Wheeler Ruml jtd7, ruml at cs.unh.edu With Thanks To NSF-IIS0812141 and DARPA CSSG N10AP20029 Jordan T. Thayer

More information

Artificial Intelligence Informed Search

Artificial Intelligence Informed Search Artificial Intelligence Informed Search Instructors: David Suter and Qince Li Course Delivered @ Harbin Institute of Technology [Many slides adapted from those created by Dan Klein and Pieter Abbeel for

More information

Heuristic Search for Planning

Heuristic Search for Planning Heuristic Search for Planning Sheila McIlraith University of Toronto Fall 2010 S. McIlraith Heuristic Search for Planning 1 / 50 Acknowledgements Many of the slides used in today s lecture are modifications

More information

Speeding up Moving-Target Search

Speeding up Moving-Target Search Speeding up Moving-Target Search Sven Koenig US S epartment skoenig@usc.edu Maxim Likhachev MU Robotics nstitute maxim+@cs.cmu.edu Xiaoxun Sun US S epartment xiaoxuns@usc.edu STRT n this paper, we study

More information

Heuristic Search and Advanced Methods

Heuristic Search and Advanced Methods Heuristic Search and Advanced Methods Computer Science cpsc322, Lecture 3 (Textbook Chpt 3.6 3.7) May, 15, 2012 CPSC 322, Lecture 3 Slide 1 Course Announcements Posted on WebCT Assignment1 (due on Thurs!)

More information

Announcements. CS 188: Artificial Intelligence

Announcements. CS 188: Artificial Intelligence Announcements Projects: Looking for project partners? --- Come to front after lecture. Try pair programming, not divide-and-conquer Account forms available up front during break and after lecture Assignments

More information

Incremental Search-Based Path Planning for Moving Target Search. Xiaoxun Sun

Incremental Search-Based Path Planning for Moving Target Search. Xiaoxun Sun Incremental Search-Based Path Planning for Moving Target Search by Xiaoxun Sun A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment

More information

CS 188: Artificial Intelligence. Recap Search I

CS 188: Artificial Intelligence. Recap Search I CS 188: Artificial Intelligence Review of Search, CSPs, Games DISCLAIMER: It is insufficient to simply study these slides, they are merely meant as a quick refresher of the high-level ideas covered. You

More information

16.410/413 Principles of Autonomy and Decision Making

16.410/413 Principles of Autonomy and Decision Making 16.410/413 Principles of Autonomy and Decision Making Lecture 17: The Simplex Method Emilio Frazzoli Aeronautics and Astronautics Massachusetts Institute of Technology November 10, 2010 Frazzoli (MIT)

More information

Planning & Decision-making in Robotics Planning Representations/Search Algorithms: RRT, RRT-Connect, RRT*

Planning & Decision-making in Robotics Planning Representations/Search Algorithms: RRT, RRT-Connect, RRT* 16-782 Planning & Decision-making in Robotics Planning Representations/Search Algorithms: RRT, RRT-Connect, RRT* Maxim Likhachev Robotics Institute Carnegie Mellon University Probabilistic Roadmaps (PRMs)

More information

Pathfinding. Artificial Intelligence for gaming

Pathfinding. Artificial Intelligence for gaming Pathfinding Artificial Intelligence for gaming Pathfinding Group AI Execution Management Strategy World Inter face Character AI Decision Making Movement Pathfinding Animation Physics Pathfinding Graphs

More information