Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
Fujita LaboratoryTokyo Institute of Technology
Basic Cases Study on Limited-Range Anisotropic Sensor in Coverage Control
Mobile Networks
Risvan Dirza09R12111
June 14, 2010(was written since March)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
2
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Game theory approach might be one solution (*)
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
3
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Game theory approach might be one solution (*)
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
4
Brief Introduction & Previous Work (1)
Background1. Search and Recovery Operations.2. Manipulation in hazardous environments.3. Surveillance4. Environmental Monitoring.
Field
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
5
START
Adequate amount of energy
Initialization of Sensor’s Position
Set of a point that will be checked
Calculating the Sensory Domain of
Agent X
The Point is in Observed Field
The Point is in Sensing Area
Calculating the Agent’s Detection Probability
Another agent (X+1) is in
Neighborhood
Calculating the Sensory Domain of
Agent (X+1)
Another agent (X+1) is in
Sensing Area
Calculating the Agent (X+1) ’s Detection Probability
Calculations of Gradient Flows Approach
Updating Status
Another agent (X+2) is in
Neighborhood
Calculating the Sensory Domain of
Agent (X+2)
Another agent (X+2) is in
Sensing Area
Calculating the Agent (X+2) ’s Detection Probability
Another agent (X+N) is in
Neighborhood
Calculating the Sensory Domain of
Agent (X+N)
Another agent (X+N) is in
Sensing Area
Calculating the Agent (X+N) ’s Detection
Probability
Y
N
END
Y
Next Point (Iteration)
N
Y
NNext Point (Iteration)
Y
N
Y
N
Y
N
Y
N
Y
N
Y
N
New Block
Brief Introduction & Previous Work (2)
There are some addition and modification inside the block
Distributed Control Algorithm
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
6
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Game theory approach might be one solution (*)
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
7
Problem Clarification (1)Previous Work
Problem : •Sensors’ movement are ineffective, •Objective Function is not maximized
High Density Region
SimulationVideo
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
8
Problem Clarification (2) Previous Work
• The energy will be decreasing proportional to the trajectory
Assume :
• The sensing radius will be proportional to the energy contained
Effect :
FROM NOWAssumptions : Finite Energy Supplies
There are 3 solutions that might overcome :1.Modifying Energy Consumption Model2.Modifying Sensor Model3.Designing Suitable Controller
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
9
Theorem & Definition (1)
Definition 1
Theorem 1 (A. Kwok et. all 2007 )
Definition 2
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
10
Theorem & Definition (2)
Theorem 2
(H. K. Khalil, Nonlinear Systems)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
11
Hypothesis (1)
Related to the problem occurred (explained before), There will be two hypothesis that should be tested :1.There is at least one variable which should be well considered to manipulate agent’s motion.2.That variable can be controlled by local information.
Gradient Flows Calculation
Local Information Data Feedback and
Calculation
+‘Old’ Next Movement ‘New’ Next Movement
The Variable
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
12
Solution (1)
1. Modifying Energy Consumption ModelPrevious Model Proposed New Model
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
13
Solution (2)
2. Modifying Sensor ModelPrevious Model
Proposed New Model
Physical Meaning
Physical MeaningToo many variablesshould be considered
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
14
Simulation I(1)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
15
Simulation II (1)
The motions are pretty smooth (No Ripple!!)
x (m)
y (m)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
16
Simulation II (2)
Positioning’s Period ;Agents are gathering around targeted area ;
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
17
Simulation III(1)
The area result are almost the as the previous report. But,The motions are pretty smooth (No Ripple!!)
x (m)
y (m)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
18
Simulation III(2)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
19
Proof (1-1)
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
20
Proof (2-1)
Physical MeaningNeed to be defined
Fact : Assigning Constant Value will decrease the objective function’s value (based on the result of the simulation)Using Theorem I Definition I and Definition II
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
21
Proof (2-2)
Using Theorem II
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
22
Failure Cases
During writing the report etc., I’m just curious of the condition of turning off one or two agents (in 800 s and 1600 s respectively). This is the result :
Although one or two agents fail, the targeted area still can be achieved.
But, don’t know how to prove it yet..
Proof of distributed algorithm.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
23
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Game theory approach might be one solution (*)
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
24
Considering Communication Cost (1)
Communication Cost mainly comes from the power/energy consumption for wireless transmission !!We consider, there are two key energy parameters :
Modeling
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
25
Considering Communication Cost (2)
New Joint Objective Function
Hypothesis
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
26
Considering Communication Cost (3)
Simulation
Reducing around 20% of Energy Cost
Detecting more events
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
27
Considering Communication Cost (4)
In spite of that, in order to perform the best position for minimizing cost. The probability of an agent to be detected by other agent is bigger. Because minimizing communication cost means indirectly push agents to be closer among them.
Minimizing the communication cost makes the sensing radius performs better for longer time. This condition make the events detection frequency is increasing.
One Agent detect other existing agents more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
28
Considering Communication Cost (4)
Proof
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
29
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Game theory approach might be one solution (*)
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
30
Avoiding Obstacles (1)
Agents will avoid any obstacles. As the compensation, it needs longer time or trajectory to maximizes the objective function
Field Model High Density Area : TargetObstacles
Hypothesis
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
31
Avoiding Obstacles (2)
Solution : New Objective Function
If we want to consider the communication cost and energy also than
To Evaluate this solution, we will show the trajectory of each agent.
The Objective Function will be :
Simulation
Objective Function
Trajectory
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
32
Outline Report
Brief Introduction and Previous WorkSelecting and Manipulating Energy Consumption Model to reduce unnecessary motion (****)Considering Communication Cost (****)Avoiding Obstacles (***)Considering Uneven Surfaces (*)Vehicle Model (**)3D Works (**)Optimum Multiple Target Achievement – Potential Game approach might be one solution ()
**** : Need to be checked*** : Need to be proven** : Has some problems in Simulation* : Need to be discussed more.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
33
Considering Uneven Surfaces (1)
Uneven Surface Model High Density Area : TargetAgents Initial Position
Hypothesis
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
34
Considering Uneven Surfaces (2)
Solution :
New Objective Function
If we want to consider the communication cost and energy also than
To Evaluate this solution, we will consider the cost of path that all of agents through. This parameter can be noticed by observing the energy status since the cost of path will decrease the amount of energy.
The Objective Function will be :
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
35
Considering Uneven Surfaces (3)
Simulation
3D Works I have problem in selecting step-size.
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
36
Considering Simplified Vehicle Model (1)
An Agent will not hit other agents by expanding the “point” into “field”.
Hypothesis :
“point” “field”
Considering Agent’s Orientation/Direction, then this model has some advantages :1.Avoiding any collision related to rotating motion (orientation)2.Easier in Programming
Simulation On Developing Progress!!
Fujita LaboratoryTokyo Institute of Technology
Tokyo Institute of Technology
37
Conclusion & Future Works