Modelling and Predicting Future Trajectories of Moving Objects in a Constrained Network

rucksackbulgeΤεχνίτη Νοημοσύνη και Ρομποτική

1 Δεκ 2013 (πριν από 3 χρόνια και 4 μήνες)

66 εμφανίσεις


Modelling and Predicting Future

Trajectories of Moving Objects in a Constrained Network



appeared in “Proceedings of the 7th International Conference on Mobile Data Management (MDM'06),japan”




Jidong Chen

Xiaofeng Meng

Yanyan Guo

S.Grumbach

Hui Sun





Information School, Renmin University of China, Beijing, China





Presented by Yanfen Xu




2


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

3


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

4


Introduction



Focus:


location modelling


future trajectory prediction



Contributions:


p
resent the graphs of cellular automata (GCA) model


propose a simulation based prediction (SP) method


experiments evaluation



5


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

6


Related Work


The modeling of MOs:


MOST model, STGS model, abstract data type


connecting road network with MOs



first in 2001, wolfson et. Al



L.Speicys: a computational data model



MODTN model



Prediction methods for future trajectories


Linear movement model


Non_linear movement models, using



quadratic predictive function,



recursive motion functions



Chebyshev polynomials







7


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

8


Graphs of Cellular Automata Model (GCA)


Modeling of the road network:



cellular automata



nodes



edges



GCA state: a mapping from cells to MOs, velocity

9


Graphs of Cellular Automata Model (GCA)


Modeling of the MOs



position can be expressed by (
startnode, endnode, measure
).



the
in
-
edge trajectory of a MO in a CA of length L:





the global trajectory of a MO in different CAs:









10


Graphs of Cellular Automata Model (GCA)


Moving rules:










P
o

11


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

12


Trajectory Prediction


The Linear Prediction (LP)



the trajectory function for an object between time
t
0

and
t
1





basic LP idea



the inadequacy of LP







13


Trajectory Prediction


The Simulation
-
based Prediction (SP)




Get the predicted positions by simulating a object






Get

the future trajectory function of a MO from the points using


regression (OLSE)







14


Trajectory Prediction


Get the
slowest and the fastest

movement function by using different P
d



Find the bounds of future positions by translating the 2 regression lines


15


Trajectory Prediction



Obtain specific future position

16


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

17


Experimental Evaluation


Datasets:


generated by: CA simulator


Brinkhoff’s Network
-
based Generator


Prediction Accuracy with Different Threshold





18


Experimental Evaluation


Prediction Accuracy with Different P
d



19


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

20


Conclusion





introduce a new model
-

GCA



propose a prediction method, based on the GCA



experiments show higher performacne than linear prediction


21


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

22


Relation to our Project


Common:


Modeling road network constrained MOs


Tracking the movement of MOs



Difference:



efficiently perform query on MOs in oracle in my project



an option to use non
-
linear predition strategy



an idea to consider the uncertainty of MO.






23


Outline


Introduction


Related Work


Graphs of Cellular Automata Model (GCA)


Trajectory Prediction


Experimental Evaluation


Conclusion


Relation to our Project


Strong and Weak Points

24


Strong and Weak Points


Strong Points




integrate traffic simulation techniques with dbs model



propose a GCA model



take correlation of MOs and stochastic hehavior into account




Weak Points



a non
-
trival prediction strategy



inconsistent position representation. (
t
i
, d
i
) and (
t
i
, l
i
)



typoes
:



25





thank you