際際滷

際際滷Share a Scribd company logo

muhsina

Download as PPTX, PDF
M
Muhsina K Shahul

The document describes a double layer ramp metering model based on adaptive neural networking. The lower model uses a backpropagation neural network to identify where traffic incident congestion occurs on an expressway. It outputs the congested section number and ramps needing control. The upper model then designs the ramp metering strategy to control ramp entry rates and optimize traffic flow. A case study showed this adaptive approach improved traffic flow over fixed-time ramp metering.

1 of 39
Download to read offline
1
DOUBLE LAYER RAMP METERING MODEL 2 PRESENTED BY, MUHSINA K SHAHUL ROLL NO:36 C.E,S7 GUIDED BY, MS.JENCY JAMES ASST.PROFESSOR DEPT OF C.E
Traffic congestion is a condition on road networks that occurs as use increases Characterized by slower speeds, longer trip times, and increased vehicular queuing Increase in traffic demand and the interaction between vehicles, slows the speed of the traffic stream 3 Traffic congestion induced by traffic accidents called "incident congestion".
Traffic congestion in Moscow( 4 INTRO (Contd) www.commons.wikimedia.org)
5 Ramp metering (www.dcsc.tudelf.nl) To alleviate traffic jam many ramp-metering measures are used on the urban expressway. INTRO (Contd)
6 RAMP- METERING Traffic signals on entrance ramp For addressing recurring freeway congestion Control the rate at which vehicles enter the mainline May be controlled locally based on time-of-day and day-of-week Traffic responsive metering -enacted based on volume, occupancy, or speed Obtained by the local freeway detection
7 Plans are stored in the controller in the same manner as surface street intersection traffic signals May also be controlled from a central system based on ramp and mainline traffic conditions Can be turned on and off from the operations center Can be controlled by central software Utilizing ramp metering algorithms to disperse traffic volumes throughout the system
8 TYPES OF RAMP METERING Simplest form of ramp metering Pre-programmed metering rate for the particular control period Based on historically averaged traffic conditions The cycle length is set and does not change Primary drawback of this metering system Fixed Time/Time of Day Metering (TOD)
9 . TYPES OF RAMP METERING (Contd) Local Traffic Responsive Metering Directly influenced by mainline traffic conditions The controller can override the set metered rates It reacts only to traffic conditions adjacent to the ramp Local traffic responsive metering is widely used
10 TYPES OF RAMP METERING (Contd) System Wide Adaptive Ramp Metering (SWARM) Adapts the local traffic responsive metering concept to a whole section of freeway An effective new tool in ramp metering Capable to adjust the metering rate of several consecutive ramp meters Based on actual traffic conditions
11 DOUBLE-LAYER RAMP METERING MODEL is an emerging technique , based on adaptive neural networking The function of the lower model -recognize where the incident congestion occurs Should identify the congestion, and design a control strategy quickly and accurately Upper model-design the ramp-metering strategy
12 CONCEPTS Models of biological neural structures The starting point is a model neuron Consists of multiple inputs and a single output. Each input is modified by a weight Multiplies with the input value The neuron will combine these weighted inputs
13 A model neuron (source: Back propagation network is the most prevalent and generalized neural network currently in use Usually fully connected CONCEPTS OF NEURAL NETWOKING (Contd..) www.cheshireeng.com)
14 Back propagation Network( CONCEPTS OF NEURAL NETWOKING (Contd..) www.cheshireeng.com)
15 Each neuron is connected to every output from the preceding layer Connected to one input from the external world if the neuron is in the first layer Output is connected to every neuron in the succeeding layer.
16 DOUBLE-LAYER RAMP-METERING MODEL AND SOLUTION DOUBLE LAYER RAMP METERING MODEL Inputs of lower model Traffic Flow Velocity Density
17 Outputs The number of section where the congestion occurs The number of ramp which should be controlled The real time traffic flow information These outputs should be transmitted to the upper model
18 The final outputs of the model Ramp-metering rate Real-time traffic-flow state after ramp controlling on the expressway
19 Basic idea of double-layer ramp-metering model, (Yihu wu et al(2014)) DOUBLE LAYER RAMP METERING MODEL (Contd)
20 Structure of expressway, (Yihu wu et al,2014)
21 The inputs are written as X= (x1,x2,x3,x4,x5,x6,x7,x8)which are shown in equation Model establishment LOWER MODEL CONGESTION IDENTIFICATION AND SOLUTION
22
23 qt , vt , ot - the traffic flow, traffic velocity, density for upper section at time t qt-1, vt-1, ot-1 is the traffic flow, traffic velocity , density for upper section at time t -1 Model establishment (Contd..) qt , vt , ot is the traffic flow, traffic velocity, density for lower section at time t
24 qt-1 is the traffic flow for lower section at time t -1 ot-1 is the density for lower section at time t -1 Model establishment (Contd..) The outputs are congestion state (y1 = 1) and no congestion state (Y2 =1 ), which could be written as Y =( 1 ,0 ) If Y =1, there is traffic incident congestion
25 The structure of the lower model is described as "3 +1" "3" means the typical 3-layer BP neural network, "1" means a transport layer to the upper model If Y =0,the output of the lower model is null; if Y =1, the dot product function would be activated Required ramp metering number would be calculated Transported to the upper model as input data Model establishment (Contd..)
26 Structure of lower model-incident congestion identification, Source (Yihu wu et al,2014) Solution Compute the response of the neural network to any set of inputs Model establishment (Contd..)
27 The initial value of the weight and number of hidden neurons were solved by particle-swarm optimization Particle swarm optimization (PSO) A computational method Iteratively improve a candidate solution With regard to a given measure of quality Solution(Contd..)
28 UPPER MODEL-RAMP METERING MODEL AND SOLUTION Model establishment The traffic capacity of the un-congested sections should be used sufficiently The expected queuing length of the ramp should be metered is a constant Expected density of each section is key density is Expressway's traffic capacity reaches its maximum
29 Objectives: Ensure the un-congested sections are used at their maximum traffic capacity Control the queuing length of the ramp as short as possible. Therefore, a function is set considering this parameters Model establishment (Contd..) Equation structure the prediction control model
30 SOLUTION The output of the lower model is transferred to the upper model For a wide range of input values, the output are obtained The final outputs are the ramp metering rate and the real time traffic flow information after the ramp controlling the expressway This is the response of the neural network
31 A CASE STUDY ON SUN-YAT-SEN FREEWAY ON-RAMP CONTROL Study conducted by Hsian Chen Lin in 1999 Compared adaptive neural network ramp system with pretimed system Pre-timed ramp control principally applied Sun-Yat-Sen freeway in Taiwan Suits in low and steady flow rather than high and rapid changed flow No effect when traffic volume approach capacity in some special festival in Taiwan
32 Average car speed dropped <40 km/hr When all drivers who work far from their hometown want go home on importance continuous holidays Study area (Hsiang-chen Lin,1999)
33 There are two lanes in freeway. Total on-ramp's length 360 m Three types flow is used to analysis their different effect under pretimed and adaptive control. Flow from 29 to 3l Jan -very high with total flow equal to 130218 vehicle/day in main lanes Flow in24 January 28 Jan and 1 Feb -middle The others are low -total flow less than 12000 vehicle/day
34 Metering rate of adaptive ramp control calculated according to predicted traffic flows Based on historical traffic flow Changes immediately when traffic situation changes
35 Table shows adaptive control method superior to pretimed control According to total delays at ramp and the number of vehicles giving up entering freeway. Source: Hsiang-chen Lin(1999) Scenarios of input data
36 RESULT Superior to pretimed ramp control The performance improves at least 69.38% Another benefit is all vehicles can enter freeway under adaptive control model But this system is more expensive than pretimed system in facilities
37 CONCLUSION The delay produced on the ramp is reduced Queuing length is small Equitableness for each driver is considered in the double-layer ramp-metering model
38 REFERENCES 1.Yihu wu et al(2014), Double-layer ramp-metering model for incident congestion on expressway, Journal of Traffic and Transportation Engineering, 2014,1(2) :129-137 2. Hsiang-chen Lin(1999), A case study on sun-yat-sen free way on-ramp control using fuzzy theory,Journal of the eastern asia society for transportation studies, vol.3, no.6, september, 1999 3.Bellemans et al.(2006),Model predictive control for ramp metering of motorway traffic;a case study.Control Engineering practice,14(7)(2006),pp.757-767. 4.Ghods et al.(2007),A genetic-fuzzy control application to ramp metering and variable speed limit control.IEEE International conference on systems,Man and cybernetics,Montreal. 5.Danzhen,G and A.Qian(2007),Application of adaptive neural network in dynamic load modeling.Proceedings of the Chinese society for Electrical Engineering,27(16)(2007),pp.31-36.
39

More Related Content

muhsina

  • 1. 1
  • 2. DOUBLE LAYER RAMP METERING MODEL 2 PRESENTED BY, MUHSINA K SHAHUL ROLL NO:36 C.E,S7 GUIDED BY, MS.JENCY JAMES ASST.PROFESSOR DEPT OF C.E
  • 3. Traffic congestion is a condition on road networks that occurs as use increases Characterized by slower speeds, longer trip times, and increased vehicular queuing Increase in traffic demand and the interaction between vehicles, slows the speed of the traffic stream 3 Traffic congestion induced by traffic accidents called "incident congestion".
  • 4. Traffic congestion in Moscow( 4 INTRO (Contd) www.commons.wikimedia.org)
  • 5. 5 Ramp metering (www.dcsc.tudelf.nl) To alleviate traffic jam many ramp-metering measures are used on the urban expressway. INTRO (Contd)
  • 6. 6 RAMP- METERING Traffic signals on entrance ramp For addressing recurring freeway congestion Control the rate at which vehicles enter the mainline May be controlled locally based on time-of-day and day-of-week Traffic responsive metering -enacted based on volume, occupancy, or speed Obtained by the local freeway detection
  • 7. 7 Plans are stored in the controller in the same manner as surface street intersection traffic signals May also be controlled from a central system based on ramp and mainline traffic conditions Can be turned on and off from the operations center Can be controlled by central software Utilizing ramp metering algorithms to disperse traffic volumes throughout the system
  • 8. 8 TYPES OF RAMP METERING Simplest form of ramp metering Pre-programmed metering rate for the particular control period Based on historically averaged traffic conditions The cycle length is set and does not change Primary drawback of this metering system Fixed Time/Time of Day Metering (TOD)
  • 9. 9 . TYPES OF RAMP METERING (Contd) Local Traffic Responsive Metering Directly influenced by mainline traffic conditions The controller can override the set metered rates It reacts only to traffic conditions adjacent to the ramp Local traffic responsive metering is widely used
  • 10. 10 TYPES OF RAMP METERING (Contd) System Wide Adaptive Ramp Metering (SWARM) Adapts the local traffic responsive metering concept to a whole section of freeway An effective new tool in ramp metering Capable to adjust the metering rate of several consecutive ramp meters Based on actual traffic conditions
  • 11. 11 DOUBLE-LAYER RAMP METERING MODEL is an emerging technique , based on adaptive neural networking The function of the lower model -recognize where the incident congestion occurs Should identify the congestion, and design a control strategy quickly and accurately Upper model-design the ramp-metering strategy
  • 12. 12 CONCEPTS Models of biological neural structures The starting point is a model neuron Consists of multiple inputs and a single output. Each input is modified by a weight Multiplies with the input value The neuron will combine these weighted inputs
  • 13. 13 A model neuron (source: Back propagation network is the most prevalent and generalized neural network currently in use Usually fully connected CONCEPTS OF NEURAL NETWOKING (Contd..) www.cheshireeng.com)
  • 14. 14 Back propagation Network( CONCEPTS OF NEURAL NETWOKING (Contd..) www.cheshireeng.com)
  • 15. 15 Each neuron is connected to every output from the preceding layer Connected to one input from the external world if the neuron is in the first layer Output is connected to every neuron in the succeeding layer.
  • 16. 16 DOUBLE-LAYER RAMP-METERING MODEL AND SOLUTION DOUBLE LAYER RAMP METERING MODEL Inputs of lower model Traffic Flow Velocity Density
  • 17. 17 Outputs The number of section where the congestion occurs The number of ramp which should be controlled The real time traffic flow information These outputs should be transmitted to the upper model
  • 18. 18 The final outputs of the model Ramp-metering rate Real-time traffic-flow state after ramp controlling on the expressway
  • 19. 19 Basic idea of double-layer ramp-metering model, (Yihu wu et al(2014)) DOUBLE LAYER RAMP METERING MODEL (Contd)
  • 20. 20 Structure of expressway, (Yihu wu et al,2014)
  • 21. 21 The inputs are written as X= (x1,x2,x3,x4,x5,x6,x7,x8)which are shown in equation Model establishment LOWER MODEL CONGESTION IDENTIFICATION AND SOLUTION
  • 22. 22
  • 23. 23 qt , vt , ot - the traffic flow, traffic velocity, density for upper section at time t qt-1, vt-1, ot-1 is the traffic flow, traffic velocity , density for upper section at time t -1 Model establishment (Contd..) qt , vt , ot is the traffic flow, traffic velocity, density for lower section at time t
  • 24. 24 qt-1 is the traffic flow for lower section at time t -1 ot-1 is the density for lower section at time t -1 Model establishment (Contd..) The outputs are congestion state (y1 = 1) and no congestion state (Y2 =1 ), which could be written as Y =( 1 ,0 ) If Y =1, there is traffic incident congestion
  • 25. 25 The structure of the lower model is described as "3 +1" "3" means the typical 3-layer BP neural network, "1" means a transport layer to the upper model If Y =0,the output of the lower model is null; if Y =1, the dot product function would be activated Required ramp metering number would be calculated Transported to the upper model as input data Model establishment (Contd..)
  • 26. 26 Structure of lower model-incident congestion identification, Source (Yihu wu et al,2014) Solution Compute the response of the neural network to any set of inputs Model establishment (Contd..)
  • 27. 27 The initial value of the weight and number of hidden neurons were solved by particle-swarm optimization Particle swarm optimization (PSO) A computational method Iteratively improve a candidate solution With regard to a given measure of quality Solution(Contd..)
  • 28. 28 UPPER MODEL-RAMP METERING MODEL AND SOLUTION Model establishment The traffic capacity of the un-congested sections should be used sufficiently The expected queuing length of the ramp should be metered is a constant Expected density of each section is key density is Expressway's traffic capacity reaches its maximum
  • 29. 29 Objectives: Ensure the un-congested sections are used at their maximum traffic capacity Control the queuing length of the ramp as short as possible. Therefore, a function is set considering this parameters Model establishment (Contd..) Equation structure the prediction control model
  • 30. 30 SOLUTION The output of the lower model is transferred to the upper model For a wide range of input values, the output are obtained The final outputs are the ramp metering rate and the real time traffic flow information after the ramp controlling the expressway This is the response of the neural network
  • 31. 31 A CASE STUDY ON SUN-YAT-SEN FREEWAY ON-RAMP CONTROL Study conducted by Hsian Chen Lin in 1999 Compared adaptive neural network ramp system with pretimed system Pre-timed ramp control principally applied Sun-Yat-Sen freeway in Taiwan Suits in low and steady flow rather than high and rapid changed flow No effect when traffic volume approach capacity in some special festival in Taiwan
  • 32. 32 Average car speed dropped <40 km/hr When all drivers who work far from their hometown want go home on importance continuous holidays Study area (Hsiang-chen Lin,1999)
  • 33. 33 There are two lanes in freeway. Total on-ramp's length 360 m Three types flow is used to analysis their different effect under pretimed and adaptive control. Flow from 29 to 3l Jan -very high with total flow equal to 130218 vehicle/day in main lanes Flow in24 January 28 Jan and 1 Feb -middle The others are low -total flow less than 12000 vehicle/day
  • 34. 34 Metering rate of adaptive ramp control calculated according to predicted traffic flows Based on historical traffic flow Changes immediately when traffic situation changes
  • 35. 35 Table shows adaptive control method superior to pretimed control According to total delays at ramp and the number of vehicles giving up entering freeway. Source: Hsiang-chen Lin(1999) Scenarios of input data
  • 36. 36 RESULT Superior to pretimed ramp control The performance improves at least 69.38% Another benefit is all vehicles can enter freeway under adaptive control model But this system is more expensive than pretimed system in facilities
  • 37. 37 CONCLUSION The delay produced on the ramp is reduced Queuing length is small Equitableness for each driver is considered in the double-layer ramp-metering model
  • 38. 38 REFERENCES 1.Yihu wu et al(2014), Double-layer ramp-metering model for incident congestion on expressway, Journal of Traffic and Transportation Engineering, 2014,1(2) :129-137 2. Hsiang-chen Lin(1999), A case study on sun-yat-sen free way on-ramp control using fuzzy theory,Journal of the eastern asia society for transportation studies, vol.3, no.6, september, 1999 3.Bellemans et al.(2006),Model predictive control for ramp metering of motorway traffic;a case study.Control Engineering practice,14(7)(2006),pp.757-767. 4.Ghods et al.(2007),A genetic-fuzzy control application to ramp metering and variable speed limit control.IEEE International conference on systems,Man and cybernetics,Montreal. 5.Danzhen,G and A.Qian(2007),Application of adaptive neural network in dynamic load modeling.Proceedings of the Chinese society for Electrical Engineering,27(16)(2007),pp.31-36.
  • 39. 39