Trajectory Simplification: On Minimizing the Direction-based Error
-
Upload
mervyn-dennis -
Category
Documents
-
view
229 -
download
0
description
Transcript of Trajectory Simplification: On Minimizing the Direction-based Error
![Page 1: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/1.jpg)
Trajectory Simplification: On Minimizing the Direction-based ErrorCheng Long, Hong Kong University of Science and
TechnologyRaymond Chi-Wing Wong, Hong Kong University of Science
and TechnologyH. V. Jagadish, University of Michigan
1
![Page 2: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/2.jpg)
Sample A Movement
2
Position (8:01AM)
Position (8:02AM)
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Position(8:09AM)
Position(8:10AM)
10 positionsPosition (8:03AM)
Consider the case where - we have a moving object o - we want to sample the positions of o every one
minute
![Page 3: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/3.jpg)
Trajectory
3
Segment
Segmentp1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Segment
Segment
Segment9 segments
10 positions
![Page 4: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/4.jpg)
Trajectory
4
Trajectory
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
9 segments
10 positions
![Page 5: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/5.jpg)
Trajectory
5
Raw trajectory data is usually very large
For an example,- 10,000 taxis- Sampling rate: 5s- positions per day!!
Issue 1: Storing all sampled positions incurs a very high space costIssue 2: Query processing big trajectory data incurs high time cost
Trajectory Simplificatio
nDrop some positions
As a result, only a portion of the positions is kept
![Page 6: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/6.jpg)
Trajectory Simplification
6
Trajectory Simplification:Drop some positions
Trajectory
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Suppose p2, p4, p5, p7, p8, p9 are dropped
Trajectory6 positions to be dropped
![Page 7: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/7.jpg)
Trajectory Simplification
7
Trajectory Simplification:Drop some positions
Simplified trajectory
p1
p3
p6
p10
![Page 8: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/8.jpg)
Trajectory Simplification
8
Trajectory Simplification:Drop some positions
Trajectory
Trajectory6 positions to be dropped
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Suppose p2, p3, p5, p6, p7, p9 are dropped
![Page 9: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/9.jpg)
Trajectory Simplification
9
Trajectory Simplification:Drop some positions
Another Simplified trajectory
p1
p4
p8
p10
![Page 10: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/10.jpg)
Trajectory Simplification
10
Trajectory Simplification:Drop some positions
Depending which positions to be dropped, it returns different simplified trajectories
Which positions should be dropped?
Our Idea:Drop the positions such that the “direction information” of the original trajectory is preserved
Direction-Preserving Trajectory Simplification (DPTS)
![Page 11: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/11.jpg)
Direction-Preserving Trajectory Simplification (DPTS)
11
Direction-Preserving Trajectory Simplification (DPTS)
What is “direction
information”?
Why to preserve
“direction information”?
How to preserve
“direction information”?
![Page 12: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/12.jpg)
What Is “Direction Information”?
12
Trajectory
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
![Page 13: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/13.jpg)
What Is “Direction Information”?
13
Direction information
![Page 14: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/14.jpg)
Direction-Preserving Trajectory Simplification (DPTS)
14
Direction-Preserving Trajectory Simplification (DPTS)
What is “direction
information”?
Why to preserve
“direction information”?
How to preserve
“direction information”?
![Page 15: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/15.jpg)
Why To Preserve “Direction Information”?
15
Applications That Use “Direction Information”Trajectory
Query Processing
Trajectory Mining
Map Matchin
g Trajectory
Clustering
Trajectory Classificati
on
Trajectory Outlier Detectio
nLee et al. SIGMOD’07, Hung et al. VLDBJ’11
Lee et al. VLDB’08
Brakatsoulas et al. IDEAS’ 04, Pelekis et al.
TIME’07
Brakatsoulas et al.
VLDB’05 Lee et al. ICDE’08
![Page 16: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/16.jpg)
Cluster 2Cluster 1
Trajectory Clustering
16
Trajectory Clustering:Partition a set of trajectories into several clusters such that
- trajectories in the same cluster are similar- trajectories in different clusters are dissimilar
T1 T2 T3 T4
T1 and T2 are similar
T3 and T4 are similar
![Page 17: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/17.jpg)
Trajectory Clustering
17
Cluster 2Cluster 1T1 T2 T3 T4
T1 and T2 have similar “direction information”
T3 and T4 have similar “direction information”
Trajectory
ClusteringLee et al.
SIGMOD’07, Hung et al. VLDBJ’11
![Page 18: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/18.jpg)
Why To Preserve “Direction Information”?
18
Applications That Use “Direction Information”Trajectory
Query Processing
Trajectory Mining
Map Matchin
g Trajectory
Clustering
Trajectory Classificati
on
Trajectory Outlier Detectio
nLee et al. SIGMOD’07, Hung et al. VLDBJ’11
Lee et al. VLDB’08
Brakatsoulas et al. IDEAS’ 04, Pelekis et al.
TIME’07
Brakatsoulas et al.
VLDB’05 Lee et al. ICDE’08Existing trajectory simplification methods:
- Pay NO attention on preserving the “direction information”
- Preserve the “position information” in most cases
![Page 19: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/19.jpg)
Direction-Preserving Trajectory Simplification (DPTS)
19
Direction-Preserving Trajectory Simplification (DPTS)
What is “direction
information”?
Why to preserve
“direction information”?
How to preserve
“direction information”?
![Page 20: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/20.jpg)
How To Preserve “Direction Information”?
20
Original trajectory: T Simplified trajectory: T’
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Direction-based Error Measurement
![Page 21: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/21.jpg)
p4
p5
p3
p6p1 p2
p3
p6
p7
p8
p9
p10
How To Preserve “Direction Information”?
21
Direction-based Error Measurement
0
𝜋 /2
𝜋3𝜋 /2 0
𝜋 /2
𝜋3𝜋 /2
Error of segment p1-p3
Error of segment p3-p6
Error of segment p6-p10
Error of T’ = the max of its segments’ errors
0
𝜋 /2
𝜋3𝜋 /2
![Page 22: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/22.jpg)
How To Preserve “Direction Information”?
22
Problem (Min-Size):Given:
- A trajectory T;- An error tolerance \eps;
Goal: A simplified trajectory T’ s.t.
- T’ has its error at most \eps;
- T’ has the fewest positions
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.
- T’ has at most W positions;
- T’ has the smallest error
VLDB 2013
Suitable when the knowledge of \eps is available
Suitable when the knowledge of \eps is NOT available, but a budget is
![Page 23: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/23.jpg)
How To Preserve “Direction Information”?
23
Algorithms For Min-ErrorError-Search
Span-Search
DP
Exact Algorithms
Approximate Algorithm
![Page 24: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/24.jpg)
DP: A Sub-Problem Optimality Property
24
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Problem Instance 1:- Trajectory: p1-p2-…-
p10- Budget: 4
p1-p3-p6-p10 is the optimal solution for Problem Instance 1
Problem Instance 2:
![Page 25: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/25.jpg)
DP: A Sub-Problem Optimality Property
25
p7
p8
p9
p10
p1 p2
p3
p4
p5 p6
Problem Instance 2:- Trajectory: p1-p2-…-
p6- Budget: 3
Problem Instance 1:- Trajectory: p1-p2-…-
p10- Budget: 4
p1-p3-p6-p10 is the optimal solution for Problem Instance 1
Sub-problem of Problem Instance 1
p1-p3-p6 is the optimal solution for Problem Instance 2
Could be verified by contradiction
![Page 26: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/26.jpg)
How To Preserve “Direction Information”?
26
Algorithms For Min-ErrorError-Search
Span-Search
DP
Exact Algorithms
Approximate Algorithm
Time: O(W n3)Space: O(n2)
![Page 27: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/27.jpg)
Feasible simplification:A simplified trajectory with at most W positions
Error-Search: A Rough Idea
27
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
All possible “feasible simplifications”:
p1-p4-p7-p10
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.- T’ has at most W positions;- T’ has the smallest error
Goal: - A feasible simplification with
the smallest error
W = 4
![Page 28: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/28.jpg)
Error-Search: A Rough Idea
28
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Feasible simplification:A simplified trajectory with at most W positions
All possible “feasible simplifications”:
p1-p4-p7-p10P1-p4-p6-p10
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.- T’ has at most W positions;- T’ has the smallest error
Goal: - A feasible simplification with
the smallest error
W = 4
![Page 29: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/29.jpg)
Error-Search: A Rough Idea
29
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Feasible simplification:A simplified trajectory with at most W positions
All possible “feasible simplifications”:
p1-p4-p7-p10P1-p4-p6-p10p1-p3-p6-p10
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.- T’ has at most W positions;- T’ has the smallest error
Goal: - A feasible simplification with
the smallest error
W = 4
![Page 30: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/30.jpg)
Error-Search: A Rough Idea
30
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Feasible simplification:A simplified trajectory with at most W positions
All possible “feasible simplifications”:
p1-p4-p7-p10P1-p4-p6-p10p1-p3-p6-p10…
Corresponding errors (in radians):
1.3051.3050.785…
The smallest error
Optimal solution
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.- T’ has at most W positions;- T’ has the smallest error
Goal: - A feasible simplification with
the smallest error
W = 4
![Page 31: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/31.jpg)
Error-Search: A Rough Idea
31
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
Feasible simplification:A simplified trajectory with at most W positions
All possible “feasible simplifications”:
p1-p4-p7-p10P1-p4-p6-p10p1-p3-p6-p10…
Corresponding errors (in radians):
1.3051.3050.785…
Size: O( nW) Size: bounded by O( n2)
Multiple simplifications share the same error
O(n2) possible segments in a simplified trajectory
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.- T’ has at most W positions;- T’ has the smallest error
Goal: - A feasible simplification with
the smallest error
W = 4
![Page 32: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/32.jpg)
How To Preserve “Direction Information”?
32
Algorithms For Min-ErrorError-Search
Span-Search
DP
Exact Algorithms
Approximate Algorithm
Time: O(W n3)Space: O(n2)
Time: O(C n2 log n)Space: O(n2)
2-factor approximation
Time: O(n log2 n)Space: O(n)
![Page 33: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/33.jpg)
Span-Search: The Main Idea
33
Original trajectory: T Simplified trajectory: T’
p1 p2
p3
p4
p5 p6
p7
p8
p9
p10
The “span” of a simplified trajectory
![Page 34: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/34.jpg)
p4
p5
p3
p6p1 p2
p3
p6
p7
p8
p9
p10
Span-Search: The Main Idea
340
𝜋 /2
𝜋3𝜋 /2 0
𝜋 /2
𝜋3𝜋 /2
Span of p1-p2-p3
Span of p3-p4-p5-p6
Span of p6-p7-p8-p9-p10
Span of T’ = the max of its sub-trajectories’ spans
The “span” of a simplified trajectory
0
𝜋 /2
𝜋3𝜋 /2
![Page 35: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/35.jpg)
Span-Search: The Main Idea
35
A Property: A simplified trajectory has its error
- at most its span and- at least half its span
0
𝜋 /2
𝜋3𝜋 /2
0
𝜋 /2
𝜋3𝜋 /2
Error of segment p1-p3
Error of segment p3-p6 0
𝜋 /2
𝜋3𝜋 /2
Error of segment p6-p10
0
𝜋 /2
𝜋3𝜋 /2 0
𝜋 /2
𝜋3𝜋 /2
Span of p1-p2-p3
Span of p3-p4-p5-p6
0
𝜋 /2
𝜋3𝜋 /2
Span of p6-p7-p8-p9-p10
![Page 36: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/36.jpg)
Span-Search: The Main Idea
36
The feasible simplification with the
smallest span
The feasible simplification with the
smallest error
A trajectory
Error_1 ½ Span_1
Span-Search
DP Error-Search
Problem (Min-Error):Given:
- A trajectory T;- A budget W;
Goal: A simplified trajectory T’ s.t.
- T’ has at most W positions;
- T’ has the smallest error
Feasible simplificationTime: O(W
n3) or O(C n2 log n)Space: O(n2)
Span_1Error_1 Span_2Error_2
2-factor approximation Time: O(n log2
n)Space: O(n)
Span_1 Span_2 Span_2 Error_2Error_1 ½ Error_2
![Page 37: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/37.jpg)
Experiments: Settings Datasets:
Geolife and T-Drive Algorithms
Exact: DP and Error-Search Approximate: Span-Search and Douglas-Peucker
Experiments DPTS vs. Wavelet Transformation Performance Studies
37
# of trajectories
Total # of
positions
Average # of
positions / trajectory
Geolife 18K 25M 1KT-Drive 10K 18M 2K
Most well-known trajectory simplification algorithm
![Page 38: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/38.jpg)
Experiments: Min-Error vs. Wavelet Transformation
38
ErrorBudget W (in terms of %)
![Page 39: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/39.jpg)
Experiments: Performance Studies on Exact Algorithms (1)
39
Running time# of positions in the
trajectory
![Page 40: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/40.jpg)
Experiments: Performance Studies on Exact Algorithms (2)
40
Running timeBudget W (in terms of
%)
![Page 41: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/41.jpg)
Experiments: Performance Studies on Approximate Algorithms
41
Appro. factorBudget W (in terms of
%)
Run. timeBudget W (in terms of %)
![Page 42: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/42.jpg)
Conclusion A New Problem: Min-Error Two Exact Algorithms: DP and Error-Search One 2-factor Approximate Algorithm: Span-
Search Experiments
42
![Page 43: Trajectory Simplification: On Minimizing the Direction-based Error](https://reader036.fdocuments.in/reader036/viewer/2022062412/5a4d1aee7f8b9ab05997cbb5/html5/thumbnails/43.jpg)
Q & A
43