CN112231870B - Intelligent generation method for railway line in complex mountain area - Google Patents
Intelligent generation method for railway line in complex mountain area Download PDFInfo
- Publication number
- CN112231870B CN112231870B CN202011011062.4A CN202011011062A CN112231870B CN 112231870 B CN112231870 B CN 112231870B CN 202011011062 A CN202011011062 A CN 202011011062A CN 112231870 B CN112231870 B CN 112231870B
- Authority
- CN
- China
- Prior art keywords
- line
- agent
- length
- cost
- ith
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000002787 reinforcement Effects 0.000 claims abstract description 5
- 239000003795 chemical substances by application Substances 0.000 claims description 82
- 238000010276 construction Methods 0.000 claims description 44
- 230000007704 transition Effects 0.000 claims description 32
- 230000009471 action Effects 0.000 claims description 29
- 238000005457 optimization Methods 0.000 claims description 16
- 230000007613 environmental effect Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000004083 survival effect Effects 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000013461 design Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 150000001875 compounds Chemical class 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 239000000463 material Substances 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000010187 selection method Methods 0.000 description 4
- 230000002068 genetic effect Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013440 design planning Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Computer Hardware Design (AREA)
- Educational Administration (AREA)
- Databases & Information Systems (AREA)
- Quality & Reliability (AREA)
- Data Mining & Analysis (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Operations Research (AREA)
- Medical Informatics (AREA)
- Remote Sensing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention relates to the technical field of railway lines, in particular to an intelligent generation method of a railway line in a complex mountain area, which comprises the following steps: firstly, building Environment; secondly, setting various attributes of Environment; thirdly, determining the optimal path of the line by adopting the DDPG in the reinforcement learning; 3.1, creating a Memory Buffer; 3.2, updating the DDPG structure parameters by using the Memory Buffer storage content; 3.3, optimizing the railway line path by using the DDPG until the line path is converged. The invention can greatly save manpower and material resources and effectively improve the efficiency and the level of the line selection of the mountain railway.
Description
Technical Field
The invention relates to the technical field of railway lines, in particular to an intelligent generation method of a railway line in a complex mountain area.
Background
The design and planning of the line are a tap and a foundation of railway construction, and are also a core work with wide related range and strong systematicness in the railway construction. The trend of the railway line directly influences the difficulty degree of project engineering, the size of engineering investment and the safety of construction and operation. Therefore, the determination of the optimal route needs to consider not only a series of natural factors such as geological topography in the railway passing area, but also other limiting conditions such as the existing railway, the road, historical historic sites, ecological environment protection areas and the like in the target area. In summary, the railway route selection problem is an optimization decision problem involving multiple limiting factors, and the optimization process is tedious and needs to consume a large amount of manpower and time to complete the optimization of each design variable.
At present, with the vigorous construction of the hidden line in the river, the complex terrain environment in the western mountainous area also brings new challenges to the reasonable arrangement of structures such as bridges, tunnels and roadbeds. Therefore, the problems can be solved by means of gradually developed artificial intelligence and computer application technology, economy and reasonability can be automatically generated, and the efficiency and the level of the line selection of the mountain railway can be effectively improved by the railway line meeting various constraints.
Many scholars at home and abroad adopt corresponding computer optimization algorithms to plan and design railway lines. Easa and the like optimize the construction cost of the railway line by adopting an enumeration method, fully consider the geometric constraint influence of the three-dimensional space line in the line optimization process, but the method has large calculation amount and cannot ensure whether the searched scheme is globally optimal or not. Hogan et al optimize a planar line by a dynamic programming method, and the method recurs an optimization target by a state transfer function to obtain a cost value of a total objective function. Jong of Maryland university utilizes a Genetic Algorithm (GA) to optimize highway lines for the first time, and the method takes plane intersection points, vertical section intersection points and plane circular curve radiuses as optimization variables, calculates the fitness (optimization target) of each generation of population through operations such as selection, intersection, variation and the like, and screens line scheme individuals according to the fitness to further obtain a line scheme with high fitness. And then, Schonfeld combines a GIS (geographic information system) with a genetic algorithm, quantifies the influence of the environment on line selection through a corresponding calculation formula, and optimizes and compares the construction cost of the railway line with the traction power performance of the train. In recent years, the scholars of the university of the middle and south have made great progress in the railway route selection in mountainous areas. A Puhao professor team provides a mountain area route optimization method based on a method of combining an improved distance transformation algorithm (DT) with a Genetic Algorithm (GA), a particle swarm algorithm (PSO) and the like, so that the constraint condition and the construction cost of the route are fully considered, factors such as geological disasters and environmental damage are quantized to serve as an optimization target of the mountain area route, and a more economic, safer and more reasonable mountain area route scheme is solved. However, the method adopted at home and abroad at present is complicated in process, generally, the line trend needs to be generated firstly, then fine tuning optimization is carried out on the line, the methodology is laggard, and the latest optimization theory in the field of artificial intelligence is not tried.
Disclosure of Invention
The invention provides an intelligent complex mountain railway line generation method which can overcome some or some defects in the prior art.
The invention discloses an intelligent generation method of a complex mountain railway line, which comprises the following steps:
firstly, building Environment;
1.1, obtaining design circuit information of a target area;
1.2, dividing a research area into a plurality of grids by utilizing a Geographic Information System (GIS);
secondly, setting various attributes of Environment;
2.1, setting the State State of the Agent, namely the existing line output by the Agent of the Agent in the Environment;
the circuit includes N times of State transition steps in the optimization process, namely transition step, and the State Space is S when the ith step is finished i :
S i ={[x i ,y i ,h i ] T |x i ∈[0,W],y i ∈[0,H]};
Wherein i is 1,2, …, N; x is the number of i And y i Represents a plane coordinate, h i The elevation of the agent at the end of the ith step; w and H are the maximum width and height of the target area grid; when i is 1, the agent is at the starting point of the line;
2.2, setting an Action of the Agent, namely the direction of the next spatial line output by the Agent;
each transition step is done by using an Action, and the Action space a is expressed as follows:
A={[Δx i ,Δy i ,G i ] T |Δx i ∈[0,W],Δy i ∈[0,H],G i ∈[-G max ,G max ]};
in the formula,. DELTA.x i And Δ y i The amount of movement of the horizontal and vertical coordinates of the plane, G, when the agent takes action i Is in a state S i And S i+1 The vertical section gradient between the two sections is required to meet the vertical section limit condition: g i ≤G max ,G i Is set to [ -G ] max ,G max ];
According to S i 、S i+1 And A i BetweenRelation of (1), state of the next step S i+1 The calculation is as follows:
S i+1 =[x i+1 ,y i+1 ,h i+1 ]=[x i +Δx i ,y i +Δy,h i +G i ×l i ];
in the formula (I), the compound is shown in the specification,is S i 、S i+1 D is the side length of the grid;
2.3, Reward of the Agent, namely after the Agent outputs the line trend, the Environment gives feedback to the Agent;
by taking action A i State of agent from S i Transition to S i+1 At this point, the Environment will give the agent a reward R i ,R i The expression of (a) is as follows:
in the formula (I), the compound is shown in the specification,andthe evaluation indexes of the unit cost, the survival state and the distance line terminal point of the line are respectively. u. of c ,u s And u d The weight coefficients of the three components;
thirdly, determining the optimal path of the line by adopting the DDPG in the reinforcement learning;
3.1 creating a Memory Buffer to store the slave S i Transition to S i+1 The transition step comprises the current state of the agent, the action in the state, the reward given to the agent by the environment after taking the action and the next state reached by the agent after taking the action;
3.2, updating DDPG structure parameters by using the storage content of Memory Buffer, randomly selecting a plurality of transition steps from the Memory Buffer for training, firstly updating Actor-Net in the Main-Net by using a calculated policy gradient by using a random gradient descent method, and then calculating a TD Error value by using two neural networks in the Target Net for updating critical-Net in the Main-Net;
3.3, optimizing the railway line path by using the DDPG until the line path is converged.
Preferably, the designing the routing information includes: elevation information of a target area, coordinates of a starting point and a middle point of a line, forbidden zone information, constraint conditions and cost standard information.
Preferably, the constraints include: maximum limit gradient G max Adjacent gradient algebraic difference delta G max Minimum curve length L Cmin Minimum clip line length L Tmin Minimum slope length L Smin Maximum allowable bridge height H Bmax And maximum allowable tunnel length L Tmax (ii) a The fee standard information includes: bridge per linear meter cost U Bi Cost of each linear meter of tunnel U Ti And fill out square unit price U Fi 、U Ci 。
Preferably, the state of the agent is subject to various constraints, including plane constraints, vertical section constraints, building structure constraints, and other constraints.
Preferably, satisfying the plane restriction condition means:
after i +1 state conversion steps, the ith rotation angle of the line is as follows:
(a) minimum plane circular curve length (L) Cmin ) It should satisfy:
L Cmin -α i R i ≤0;
in the formula, R i In order to satisfy the radius of the circular curve of the plane at the ith corner after i +1 state transition steps: r min -R i ≤0;
(b) Minimum clip line length (L) between two planar circular curves Tmin ) It should satisfy:
L Tmin -L Ti ≤0;
wherein L is Ti The calculation formula for the clip line length is as follows:
the condition that the longitudinal section limit value is met is as follows:
the longitudinal section limiting conditions mainly comprise: maximum limit gradient (G) max ) Minimum slope segment length (L) Smin ) And adjacent gradient algebraic difference (Δ G) max ):
(a) Vertical gradient:
(b) length of slope section:
(c) algebraic difference of adjacent slope sections:
|G i+1 -G i |≤ΔG max ;
satisfying the building structure limitation condition means that:
after each transition step, the intelligent agent automatically arranges bridges, tunnels and roadbed sections according to corresponding line bridge boundary filling height and line tunnel boundary excavation depth, wherein the height (H) of the bridge from the ground Bi ) Should not exceed the maximum allowable bridge height (H) Bmax ) Full length of a single tunnel (L) Tui ) Should not exceed the maximum allowable tunnel length (L) Tmax ) Namely:
H Bi ≤H Bmax ;
L Tui ≤L Tmax ;
satisfying other constraints means:
the line cannot cross the environmental protection area and the historical trails.
As a preference, the first and second liquid crystal compositions are,andthe method comprises the following steps:
Line construction costs include bridge construction costs (C) B ) Cost of tunnel construction (C) T ) Fill and dig square charge (C) E ) Environmental protection costs (C) I ) And linear cost (C) L );
(a) Bridge construction costs (C) B ):
In the formula, n is the number of full-line bridges; u shape Bi The unit construction cost ([ gamma ]/] of the ith bridge is calculated; l is Bi The length of the ith bridge; c Ai Construction costs ('this') of the i-th bridge abutment;
(b) cost of tunnel construction (C) T ):
In the formula, n is the number of the full-line tunnels; u shape Ti The unit construction cost ([ gamma ] for the ith tunnel; l is a radical of an alcohol Tui The length of the ith tunnel is taken as the length of the ith tunnel; c Pi The construction cost (rah) of the ith tunnel portal;
(c) fill and dig square charge (C) E ):
The cross-sectional area of the subgrade section can be calculated as follows:
A=2(W s +Δh×i)×Δh;
in the formula, W s The width of the roadbed surface; delta h is the roadbed filling and excavating height; i is the slope of the roadbed side slope;
further, fill and dig square fee is calculated as:
in the formula of U Fi And U Ci Respectively the ith section filling and excavating unit cost ([ gamma/m ] 3 ) M and n are the number of filling and digging respectively; a. the i Is the ith cross-sectional area (m) 2 );L i Is the ith section length (m);
(d) cost of environmental protection (C) I ):
In the formula of U i Is the unit fine ([ gamma/m ]) of the line passing through the environment protection area 2 );A Pi Area (m) occupied by a line crossing an environmentally protected area 2 );
(e) Linear cost (C) L ):
Linear costs are costs associated with track length, including track laying and electrical utility costs:
C L =U L ×L;
in the formula of U L The unit construction cost ([ m ]) for the linear expense; l is the total line length;
Based on the above fee criteria information, the state of the agent is determined by the slave S i Transition to S i+1 The unit construction cost of time is:
Agent state driven by S i Transition to S i+1 In time, all constraints can be met, and the environment will give the agent a positive survival rewardOn the contrary, if the intelligent agent can not meet all the limiting conditions, the index is a negative value;
The calculation is as follows:
in the formula (d) 1 Is the diagonal length of the target area; d is a radical of 2 The distance from the current position of the intelligent agent to the line terminal point.
Preferably, in step 3.2, the structural parameters in Target Net are not updated directly, and after every multiple iteration steps, two neural network parameters in Main-Net are copied to Target Net to realize updating.
The traditional railway route selection method is also called manual route selection in a certain sense, and is characterized in that route selection designers evaluate, analyze and compare and select a plurality of railway schemes according to past knowledge accumulation and working experience, and finally determine an optimal route. However, due to time and resource constraints, the designer's preferred routes are always limited, and many potentially viable routes are easily ignored. The intelligent route selection method provided by the invention can greatly save manpower and material resources, and effectively improve the efficiency and level of route selection of mountainous railways.
Drawings
Fig. 1 is a flowchart of an intelligent complex mountain railway route generation method in embodiment 1;
fig. 2 is a schematic diagram of meshing of topographic information in embodiment 1;
FIG. 3 is a diagram illustrating the environment and attributes of line selection optimization in embodiment 1;
FIG. 4 is a schematic line shape of a planar wiring in example 1;
FIG. 5 is a schematic diagram showing the operation in example 1;
FIG. 6 is a schematic diagram of the structure of DDPG in example 1;
FIG. 7 is a comparison between the method of the present embodiment and the manual route selection method in example 1.
Detailed Description
For a further understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings and examples. It is to be understood that the examples are illustrative of the invention and not limiting.
Example 1
As shown in fig. 1, the present embodiment provides an intelligent generation method for a complex mountain railway line, which includes the following steps:
firstly, building Environment;
1.1, obtaining design circuit information of a target area;
1.2, dividing a research area into a plurality of grids by utilizing a Geographic Information System (GIS); as shown in fig. 2, the side length of the grid varies according to the accuracy of the data, and is typically 30m, and each grid contains elevation information of the current grid;
secondly, setting various attributes of Environment;
the method is based on Reinforcement Learning, and mainly comprises the following concepts and attributes, as shown in fig. 2:
1) environment: the method comprises the steps of obtaining target area topographic information, line design information, cost standard information, various constraint conditions and the like;
2) agent of the Agent: the intelligent line selection system can be regarded as a designer of line selection or an intelligent program of line selection and is responsible for outputting an optimized line;
3) state: existing lines output by the agent in the Environment;
4) reward: after the intelligent agent outputs the route, the environment gives feedback to the intelligent agent; such as: the cost of the intelligent agent output line, whether constraint conditions are met, and the like;
5) and (4) Action: outputting the next spatial line trend by the intelligent agent;
2.1, setting the State State of the Agent, namely the existing line output by the Agent of the Agent in the Environment;
the circuit includes N times of State transition steps in the optimization process, i.e. transition step, then the State Space is S when the ith step is finished i :
S i ={[x i ,y i ,h i ] T |x i ∈[0,W],y i ∈[0,H]};
Wherein i is 1,2, …, N; x is the number of i And y i Represents a plane coordinate, h i The elevation of the agent at the end of the ith step; w and H are the maximum width and height of the target area grid; when i is 1, the agent is at the starting point of the line;
2.2, setting an Action of the Agent, namely the direction of the next spatial line output by the Agent;
as shown in fig. 5, (a) in fig. 5 is a schematic plane operation diagram, and (b) in fig. 5 is a schematic longitudinal section operation diagram; each transition step is done by using an Action, and the Action space a is expressed as follows:
A={[Δx i ,Δy i ,G i ] T |Δx i ∈[0,W],Δy i ∈[0,H],G i ∈[-G max ,G max ]};
in the formula,. DELTA.x i And Δ y i The amount of movement of the horizontal and vertical coordinates of the plane, G, when the agent takes action i Is in a state S i And S i+1 The vertical section gradient between the two sections is required to meet the vertical section limit condition: g i ≤G max ,G i Is set to [ -G ] max ,G max ];
According to S i 、S i+1 And A i Relation between, state S of the next step i+1 The calculation is as follows:
S i+1 =[x i+1 ,y i+1 ,h i+1 ]=[x i +Δx i ,y i +Δy,h i +G i ×l i ];
in the formula (I), the compound is shown in the specification,is S i 、S i+1 D is the grid side length;
2.3, Reward of the Agent, namely after the Agent outputs the line trend, the Environment gives feedback to the Agent;
by taking action A i State of agent from S i Transition to S i+1 At this point, the Environment will give the agent a reward R i ,R i The expression of (a) is as follows:
in the formula (I), the compound is shown in the specification,andthe evaluation indexes of the unit cost, the survival state and the distance line terminal point of the line are respectively. u. of c ,u s And u d The weight coefficients of the three are;
thirdly, determining the optimal path of the line by adopting the DDPG in the reinforcement learning; the DDPG includes four neural networks, and the structure thereof is shown in fig. 6, and can be divided into two types (Actor-Net, critical-Net), where the input of Actor-Net is the state (state) of the agent, and the output is the action (action); the Critic-Net is responsible for evaluating the quality of the action adopted by the intelligent agent in the current state, and the action adopted by the intelligent agent is more favorable when the output value (policy gradient) is larger; in addition, Actor-Net 'and Critic-Net' in Target-Net in FIG. 6 are identical to the two neural network structures in Main-Net, but the updating manner is different.
3.1 creating a Memory Buffer to store the slave S i Transition to S i+1 The transition step includes the current state of the agent, the action in this state, the reward given to the agent by the environment after taking this action, and the next state reached by the agent after taking this action;
3.2, updating DDPG structure parameters by using the storage content of Memory Buffer, randomly selecting a plurality of transition steps from the Memory Buffer for training, firstly updating Actor-Net in the Main-Net by using a calculated policy gradient by using a random gradient descent method, and then calculating a TD Error value by using two neural networks in the Target Net for updating critical-Net in the Main-Net;
3.3, optimizing the railway line path by using the DDPG until the line path is converged.
In this embodiment, designing the line information includes: elevation information of a target area, coordinates of a starting point and a middle point of a line, forbidden zone information, constraint conditions and cost standard information.
In this embodiment, the constraint conditions include: maximum limit gradient G max Adjacent gradient algebraic difference delta G max Minimum curve length L Cmin Minimum clip line length L Tmin Minimum slope length L Smin Maximum allowable bridge height H Bmax And maximum allowable tunnel length L Tmax (ii) a The fee standard information includes: bridge per linear meter cost U Bi Cost of each linear meter of tunnel U Ti And fill out square unit price U Fi 、U Ci 。
In this embodiment, the state of the agent needs to satisfy various restriction conditions, including plane restriction conditions, longitudinal section restriction conditions, building structure restriction conditions, and other restriction conditions.
In this embodiment, satisfying the plane constraint condition means:
after i +1 state conversion steps, the ith rotation angle of the line is as follows:
(a) minimum plane circular curve length (L) Cmin ) It should satisfy:
L Cmin -α i R i ≤0;
in the formula, R i In order to satisfy the radius of the circular curve of the plane at the ith corner after i +1 state transition steps: r min -R i ≤0;
(b) Minimum clip line length (L) between two planar circular curves Tmin ) It should satisfy:
L Tmin -L Ti ≤0;
wherein, referring to FIG. 4, L Ti The calculation formula for the clip line length is as follows:
the condition that the longitudinal section limit value is met is as follows:
the longitudinal section limiting conditions mainly comprise: maximum limit gradient (G) max ) Minimum slope segment length (L) Smin ) And adjacent gradient algebraic difference (Δ G) max ):
(a) Vertical gradient:
(b) length of slope section:
(c) algebraic difference of adjacent slope sections:
|G i+1 -G i |≤ΔG max ;
satisfying the building structure limitation condition means that:
after each transition step, the intelligent agent automatically arranges bridges, tunnels and roadbed sections according to corresponding line bridge boundary filling height and line tunnel boundary excavation depth, wherein the height (H) of the bridge from the ground Bi ) Should not exceed the maximum allowable bridge height (H) Bmax ) Full length of a single tunnel (L) Tui ) Should not exceed the maximum allowed tunnel length (L) Tmax ) Namely:
H Bi ≤H Bmax ;
L Tui ≤L Tmax ;
satisfying other constraints means:
the line cannot cross the environmental protection area and the historical trails.
Line construction costs include bridge construction costs (C) B ) Cost of tunnel construction (C) T ) Fill and dig square charge (C) E ) Environmental protection costs (C) I ) And linear cost (C) L );
(a) Bridge construction costs (C) B ):
In the formula, n is the number of full-line bridges; u shape Bi The unit construction cost ([ gamma ]/] of the ith bridge is calculated; l is Bi The length of the ith bridge; c Ai Construction costs ('this') of the i-th bridge abutment;
(b) cost of tunnel construction (C) T ):
In the formula, n is the number of the full-line tunnels; u shape Ti The unit construction cost ([ gamma ] for the ith tunnel; l is Tui The length of the ith tunnel; c Pi The construction cost (rah) of the ith tunnel portal;
(c) fill and dig square charge (C) E ):
The cross-sectional area of the subgrade section can be calculated as follows:
A=2(W s +Δh×i)×Δh;
in the formula, W s The width of the roadbed surface; delta h is the roadbed filling and excavating height; i is the slope of the roadbed side slope;
further, fill and dig square fee is calculated as:
in the formula of U Fi And U Ci Respectively the ith section filling and excavating unit cost ([ gamma/m ] 3 ) M and n are the number of filling and digging respectively; a. the i Is the ith cross-sectional area (m) 2 );L i Is the ith section length (m);
(d) cost of environmental protection (C) I ):
In the formula of U i Is the unit fine ([ gamma/m ]) of the line passing through the environment protection area 2 );A Pi Area (m) occupied by a line crossing an environmentally protected area 2 );
(e) Linear cost (C) L ):
Linear costs are costs associated with track length, including track laying and electrical utility costs:
C L =U L ×L;
in the formula of U L The unit construction cost ([ m ]) for the linear expense; l is the total line length;
Based on the above fee criteria information, the state of the agent is determined by the slave S i Transition to S i+1 The unit construction cost of time is:
notably, since the aim of the method is to reduce the construction costs of the line,taking a negative value;
In consideration of the fact that mountainous area lines are difficult to meet various limiting conditions in the design planning process, the intelligent agent cannot fully explore the target area. Therefore, an index of survival status is added in the methodThe agent can reach the end point more easily;
agent state driven by S i Transition to S i+1 In time, all constraints can be met, and the environment will give the agent a positive survival rewardOn the contrary, if the intelligent agent can not meet all the limiting conditions, the index is a negative value;
The calculation is as follows:
in the formula (d) 1 Is the diagonal length of the target area; d 2 The distance from the current position of the intelligent agent to the line terminal point. It can be seen that the closer to the endpoint, the greater the reward the agent receives.
In this embodiment, in step 3.2, the structural parameters in Target Net are not updated directly, and after every multiple iteration steps, two neural network parameters in Main-Net are copied to Target Net to realize updating. In FIG. 6, the black arrow indicates the updating step of Actor-Net in Main-Net, the red arrow indicates the updating step of Critic-Net in Main-Net, and the orange arrow indicates the updating step of Target-Net.
This example compares the method with the manual line selection method, and the result is shown in fig. 7, where the line type in fig. 7(a) is the line type obtained by the method of this example, and the lower half line types from E to S in fig. 7(b) are the results of manual line selection. Table 1 shows a comparison of the economic indicators of two lines, as can be seen: the total cost of the line obtained by the method of the embodiment is obviously lower than the result of manual line selection, and the total cost of the line is reduced by 7.37 percent.
TABLE 1 comparison of economic indicators
The present invention and its embodiments have been described above schematically, without limitation, and what is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Therefore, if the person skilled in the art receives the teaching, without departing from the spirit of the invention, the person skilled in the art shall not inventively design the similar structural modes and embodiments to the technical solution, but shall fall within the scope of the invention.
Claims (6)
1. An intelligent generation method for a railway line in a complex mountain area is characterized by comprising the following steps: the method comprises the following steps:
firstly, building Environment;
1.1, obtaining design circuit information of a target area;
1.2, dividing a research area into a plurality of grids by utilizing a Geographic Information System (GIS);
secondly, setting various attributes of Environment;
2.1, setting the State State of the Agent, namely the existing line output by the Agent of the Agent in the Environment;
the circuit includes N times of State transition steps in the optimization process, namely transition step, and the State Space is S when the ith step is finished i :
S i ={[x i ,y i ,h i ] T |x i ∈[0,W],y i ∈[0,H]};
Wherein i is 1,2, …, N; x is the number of i And y i Represents a plane coordinate, h i The elevation of the agent at the end of the ith step; w and H are the maximum width and height of the target area grid; when i is 1, the agent is at the starting point of the line;
2.2, setting an Action of the Agent, namely the direction of the next spatial line output by the Agent;
each transition step is done by taking an Action, and the Action space a is represented as follows:
A={[Δx i ,Δy i ,G i ] T |Δx i ∈[0,W],Δy i ∈[0,H],G i ∈[-G max ,G max ]};
in the formula,. DELTA.x i And Δ y i The amount of movement of the horizontal and vertical coordinates of the plane, G, when the agent takes action i Is in a state S i And S i+1 The vertical section gradient between the two sections is required to meet the vertical section limit condition: g i ≤G max ,G i Is set to [ -G ] max ,G max ];
According to S i 、S i+1 And A i Relation between, state S of the next step i+1 The calculation is as follows:
S i+1 =[x i+1 ,y i+1 ,h i+1 ]=[x i +Δx i ,y i +Δy,h i +G i ×l i ];
in the formula (I), the compound is shown in the specification,is S i 、S i+1 D is the side length of the grid;
2.3, Reward of the Agent, namely after the Agent outputs the line trend, the Environment gives feedback to the Agent;
by taking action A i State of agent from S i Transition to S i+1 At this point, the Environment will give the agent a reward R i ,R i The expression of (a) is as follows:
in the formula (I), the compound is shown in the specification,andevaluation of unit cost, survival status and distance to end point of lineA price index; u. of c ,u s And u d The weight coefficients of the three are;
thirdly, determining the optimal path of the line by adopting the DDPG in the reinforcement learning;
3.1 creating a Memory Buffer to store the slave S i Transition to S i+1 The transition step comprises the current state of the agent, the action in the state, the reward given to the agent by the environment after taking the action and the next state reached by the agent after taking the action;
3.2, updating DDPG structure parameters by using the storage content of Memory Buffer, randomly selecting a plurality of transition steps from the Memory Buffer for training, firstly updating Actor-Net in the Main-Net by using a calculated policy gradient by using a random gradient descent method, and then calculating a TD Error value by using two neural networks in the Target Net for updating critical-Net in the Main-Net;
3.3, optimizing the railway line path by using the DDPG until the line path is converged.
2. The intelligent complex mountain railway line generation method according to claim 1, characterized in that: designing the routing information includes: elevation information of a target area, coordinates of a starting point and a middle point of a line, forbidden zone information, constraint conditions and cost standard information.
3. The intelligent generation method of the complex mountain railway line according to claim 2, characterized in that: the constraint conditions include: maximum limit gradient G max Adjacent gradient algebraic difference delta G max Minimum curve length L Cmin Minimum clip line length L Tmin Minimum slope length L Smin Maximum allowable bridge height H Bmax And maximum allowable tunnel length L Tmax (ii) a The fee standard information includes: bridge per linear meter cost U Bi Cost of each linear meter of tunnel U Ti And fill out square unit price U Fi 、U Ci 。
4. The intelligent generation method of the complex mountain railway line according to claim 3, characterized in that: the state of the intelligent agent needs to meet various limiting conditions, wherein the various limiting conditions comprise plane limiting conditions, longitudinal section limiting conditions, building structure limiting conditions and other limiting conditions;
satisfying the plane constraint condition means:
after i +1 state conversion steps, the ith rotation angle of the line is as follows:
(a) length L of least plane circular curve Cmin It should satisfy:
L Cmin -α i R i ≤0;
in the formula, R i After i +1 state conversion steps, the radius of the plane circular curve at the ith corner must satisfy the following conditions: r min -R i ≤0;
(b) Minimum clip line length L between two plane circular curves Tmin It should satisfy:
L Tmin -L Ti ≤0;
wherein L is Ti The calculation formula for the clip line length is as follows:
the condition that the longitudinal section limit value is met is as follows:
the longitudinal section limiting conditions mainly comprise: maximum limit gradient G max Minimum slope length L Smin And adjacent gradient algebraic difference deltag max :
(a) Vertical gradient:
(b) length of slope section:
(c) algebraic difference between adjacent slope segments:
|G i+1 -G i |≤ΔG max ;
satisfying the building structure limitation condition means that:
after each transition step, the intelligent agent can automatically arrange the bridge, the tunnel and the roadbed section according to the corresponding line bridge boundary filling height and line tunnel boundary digging depth, wherein the height H of the bridge from the ground Bi Should not exceed the maximum allowable bridge height H Bmax Full length L of a single tunnel Tui Should not exceed the maximum allowed tunnel length L Tmax Namely:
H Bi ≤H Bmax ;
L Tui ≤L Tmax ;
satisfying other constraints means:
the line cannot cross the environmental protection area and the historical trails.
5. The intelligent generation method of the complex mountain railway line according to claim 4, characterized in that:andthe method comprises the following steps:
Line construction costs include bridge construction costs C B Cost of tunnel construction C T Fill and dig square charge C E Environmental protection charge C I And linear cost C L ;
(a) Bridge construction cost C B :
In the formula, n is the number of full-line bridges; u shape Bi The unit construction cost of the ith bridge; l is Bi The length of the ith bridge; c Ai Construction cost of the i-th bridge abutment;
(b) cost of tunnel construction C T :
In the formula, n is the number of the full-line tunnels; u shape Ti The unit construction cost of the ith tunnel is saved; l is Tui The length of the ith tunnel is taken as the length of the ith tunnel; c Pi The construction cost of the ith tunnel portal;
(c) fill and dig square charge C E :
The cross-sectional area of the subgrade section can be calculated as follows:
A=2(W s +Δh×i)×Δh;
in the formula, W s The width of the roadbed surface; delta h is the roadbed filling and excavating height; i is the slope of the roadbed side slope;
further, fill and dig square fee is calculated as:
in the formula of U Fi And U Ci Respectively the unit cost of filling and digging square of the ith section, and m and n are respectively the quantity of filling and digging square; a. the i Is the ith cross-sectional area; l is i Is the length of the ith section;
(d) cost for environmental protection C I :
In the formula of U i The unit fine when the line passes through the environmental protection area; a. the Pi The area occupied by the line passing through the environmental protection area;
(e) linear cost C L :
Linear costs are costs associated with track length, including track laying and electrical utility costs:
C L =U L ×L;
in the formula of U L Unit construction cost for linear expense; l is the total line length;
Based on the above fee criteria information, the state of the agent is determined by the slave S i Transition to S i+1 The unit construction cost of time is:
Agent state driven by S i Transition to S i+1 In time, all constraints can be met, and the environment will give the agent a positive survival rewardOn the contrary, if the intelligent agent can not meet all the limiting conditions, the index is a negative value;
The calculation is as follows:
in the formula (d) 1 Is the diagonal length of the target area; d 2 The distance from the current position of the intelligent agent to the line terminal point.
6. The intelligent generation method of the complex mountain railway line according to claim 1, characterized in that: in step 3.2, the structural parameters in Target Net are not updated directly, and after every multiple iteration steps, two neural network parameters in Main-Net are copied to Target Net to realize updating.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011011062.4A CN112231870B (en) | 2020-09-23 | 2020-09-23 | Intelligent generation method for railway line in complex mountain area |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011011062.4A CN112231870B (en) | 2020-09-23 | 2020-09-23 | Intelligent generation method for railway line in complex mountain area |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112231870A CN112231870A (en) | 2021-01-15 |
CN112231870B true CN112231870B (en) | 2022-08-02 |
Family
ID=74107687
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011011062.4A Active CN112231870B (en) | 2020-09-23 | 2020-09-23 | Intelligent generation method for railway line in complex mountain area |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112231870B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113032876B (en) * | 2021-03-19 | 2022-05-24 | 中南大学 | Method and system for laying interchange channel for automatically changing existing road along railway |
CN113642852B (en) * | 2021-07-20 | 2023-09-12 | 合肥泰瑞数创科技有限公司 | Road paving cost calculation method and system based on digital twin |
CN114692352B (en) * | 2022-04-06 | 2023-02-03 | 中南大学 | Intelligent layout method for highway network in mountain railway construction |
CN114861368B (en) * | 2022-06-13 | 2023-09-12 | 中南大学 | Construction method of railway longitudinal section design learning model based on near-end strategy |
CN115809502B (en) * | 2023-02-09 | 2023-04-25 | 西南交通大学 | Mountain railway major engineering matching Intelligent planning and design method for road |
CN116757347B (en) * | 2023-06-19 | 2024-02-13 | 中南大学 | Railway line selection method and system based on deep learning |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103077282A (en) * | 2013-01-15 | 2013-05-01 | 西南交通大学 | Rapid modeling method for high-speed railway scene based on basic-element combination and multilevel semantic constraint |
WO2015044860A1 (en) * | 2013-09-24 | 2015-04-02 | Koninklijke Philips N.V. | System for optimizing workflow for efficient on-site data collection and determination of energy analysis and method of operation thereof |
CN104504476A (en) * | 2014-12-30 | 2015-04-08 | 中南大学 | Automatic complicated mountain railway line trend generating method |
CN105243233A (en) * | 2015-11-04 | 2016-01-13 | 中南大学 | Line-station collaborative optimization method for railway in complicated mountain area |
CN107194612A (en) * | 2017-06-20 | 2017-09-22 | 清华大学 | A kind of train operation dispatching method learnt based on deeply and system |
CN107273616A (en) * | 2017-06-16 | 2017-10-20 | 中南大学 | One kind is based on the high-speed railway adjacent lines graphic design method of " intersection method " |
CN109447437A (en) * | 2018-10-17 | 2019-03-08 | 中南大学 | A kind of public affairs (iron) road vertical section method for auto constructing comprising cut-fill transition |
CN109740839A (en) * | 2018-11-23 | 2019-05-10 | 北京交通大学 | Train Dynamic method of adjustment and system under a kind of emergency event |
CN109918803A (en) * | 2019-03-13 | 2019-06-21 | 中国铁路设计集团有限公司 | The flat real-time linkage design method in length and breadth of rail track |
CN110263437A (en) * | 2019-06-20 | 2019-09-20 | 中铁二院工程集团有限责任公司 | Rail link automatic design method |
CN111553467A (en) * | 2020-04-30 | 2020-08-18 | 陈永聪 | Method for realizing general artificial intelligence |
CN111582469A (en) * | 2020-03-23 | 2020-08-25 | 成都信息工程大学 | Multi-agent cooperation information processing method and system, storage medium and intelligent terminal |
CN111598311A (en) * | 2020-04-28 | 2020-08-28 | 北京交通大学 | Novel intelligent optimization method for train running speed curve |
CN111597621A (en) * | 2020-05-26 | 2020-08-28 | 西南交通大学 | Railway line double-layer optimization method based on GIS and differential evolution algorithm |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110516532B (en) * | 2019-07-11 | 2022-03-11 | 北京交通大学 | Unmanned aerial vehicle railway track line identification method based on computer vision |
CN110428117B (en) * | 2019-08-15 | 2021-10-22 | 北京交通大学 | Passenger flow accurate induction method and system under urban rail transit multi-scene |
-
2020
- 2020-09-23 CN CN202011011062.4A patent/CN112231870B/en active Active
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103077282A (en) * | 2013-01-15 | 2013-05-01 | 西南交通大学 | Rapid modeling method for high-speed railway scene based on basic-element combination and multilevel semantic constraint |
WO2015044860A1 (en) * | 2013-09-24 | 2015-04-02 | Koninklijke Philips N.V. | System for optimizing workflow for efficient on-site data collection and determination of energy analysis and method of operation thereof |
CN104504476A (en) * | 2014-12-30 | 2015-04-08 | 中南大学 | Automatic complicated mountain railway line trend generating method |
CN105243233A (en) * | 2015-11-04 | 2016-01-13 | 中南大学 | Line-station collaborative optimization method for railway in complicated mountain area |
CN107273616A (en) * | 2017-06-16 | 2017-10-20 | 中南大学 | One kind is based on the high-speed railway adjacent lines graphic design method of " intersection method " |
CN107194612A (en) * | 2017-06-20 | 2017-09-22 | 清华大学 | A kind of train operation dispatching method learnt based on deeply and system |
CN109447437A (en) * | 2018-10-17 | 2019-03-08 | 中南大学 | A kind of public affairs (iron) road vertical section method for auto constructing comprising cut-fill transition |
CN109740839A (en) * | 2018-11-23 | 2019-05-10 | 北京交通大学 | Train Dynamic method of adjustment and system under a kind of emergency event |
CN109918803A (en) * | 2019-03-13 | 2019-06-21 | 中国铁路设计集团有限公司 | The flat real-time linkage design method in length and breadth of rail track |
CN110263437A (en) * | 2019-06-20 | 2019-09-20 | 中铁二院工程集团有限责任公司 | Rail link automatic design method |
CN111582469A (en) * | 2020-03-23 | 2020-08-25 | 成都信息工程大学 | Multi-agent cooperation information processing method and system, storage medium and intelligent terminal |
CN111598311A (en) * | 2020-04-28 | 2020-08-28 | 北京交通大学 | Novel intelligent optimization method for train running speed curve |
CN111553467A (en) * | 2020-04-30 | 2020-08-18 | 陈永聪 | Method for realizing general artificial intelligence |
CN111597621A (en) * | 2020-05-26 | 2020-08-28 | 西南交通大学 | Railway line double-layer optimization method based on GIS and differential evolution algorithm |
Non-Patent Citations (3)
Title |
---|
Harshad Khadilkar.A Scalable Reinforcement Learning Algorithm for Scheduling Railway Lines.《IEEE Transactions on Intelligent Transportation Systems》.2018,727-736. * |
李懿等.基于深度强化学习的公路初始路径寻优方法.《交通科学与工程》.2020,(第1期),98-103. * |
马小云.基于BP神经网络的西南岩溶斜坡区铁路选线评价.《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》.2020,(第02期),C033-26. * |
Also Published As
Publication number | Publication date |
---|---|
CN112231870A (en) | 2021-01-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112231870B (en) | Intelligent generation method for railway line in complex mountain area | |
Li et al. | Mountain railway alignment optimization with bidirectional distance transform and genetic algorithm | |
Li et al. | Methodology for optimizing constrained 3-dimensional railway alignments in mountainous terrain | |
CN109359350B (en) | Wind power plant road intelligent design method for optimizing fine construction cost | |
Kim et al. | Improving the computational efficiency of highway alignment optimization models through a stepwise genetic algorithms approach | |
CN110569538A (en) | Parameterized modeling and design optimization method for construction access road of railway engineering | |
CN112883474B (en) | Layout method, system, terminal and readable storage medium for intelligent reconstruction of existing track line | |
CN107885960A (en) | A kind of earthwork estimation system and evaluation method based on construction road automatic route selection in wind power plant field | |
CN113032876B (en) | Method and system for laying interchange channel for automatically changing existing road along railway | |
CN102129712A (en) | Triangular net model constructing method based on multi-stratum and three-dimensional quantity of earth work | |
CN108053060B (en) | Booster station site selection system and site selection method based on automatic road line selection in wind power plant | |
CN116739519A (en) | Highway engineering construction informatization management system and method based on BIM and GIS | |
CN114820957B (en) | Three-dimensional intelligent design method for railway line-side terrace | |
Gao et al. | Low-construction-emission cross-section optimization for mountainous highway alignment designs | |
CN105243233B (en) | A kind of railway line station, complex mountainous cooperative optimization method | |
CN107908899A (en) | A kind of wire selection system and selection method for Construction of Wind Power construction road | |
Wang | Effect of genetic algorithm in optimizing deep foundation pit supporting structure | |
Hammad et al. | Bridge planning using GIS and expert system approach | |
CN117034614A (en) | Comprehensive control and early warning system for flood disasters of transformer substation | |
Gao et al. | Low-construction-emissions optimization for railway alignment fine-grained designs within densely populated urban areas | |
CN103020741B (en) | A kind of economic optimization method for designing of urban landscape rivers regulation | |
Medvedeva et al. | Routes of laying gas supply system pipeline | |
Khalaf et al. | Multi-objective groundwater management using genetic algorithms in Kerbala desert area, Iraq | |
CN104504476B (en) | A kind of complex mountainous rail track moves towards automatic generation method | |
Gao et al. | Mathematical Interpolation and Correction of Three-Dimensional Modelling of High-Speed Railway. |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |