Collaborative optimization method of transportation planning and land use and its system unit decision making

Abstract

This paper proposes a collaborative optimization method for transportation planning and land use. This method combines system dynamics models, structural equations, DEA evaluation and BP neural network technology to construct a complete transportation planning and land use collaborative optimization system. In this system, system unit determination is a key problem, which has a different relationship with each stage. This paper uses the DEMATEL model to construct a causality diagram and an incidence matrix for the system units, to identify the central influencing unit and causal influencing unit of the method of collaborative optimization of transportation planning and land use. While simplifying the model through the DEMATEL model, the macro interaction mechanism between urban traffic and land use will be clarified. Further, the extracted key influence units can be micro-decomposed to construct subsystems of micro-elements, which makes preliminary preparation for the analysis of static and dynamic integration model.

Keywords

Transportation planning land use system unit DEMATEL

1. Introduction

As two relatively independent fields, transportation planning and land use have an interactive relationship between cyclical effects and “positive and negative” feedback [1]. Land use determines the needs of the transportation system, and the planning and adjustment of the transportation system in turn restricts the intensity and type of land use. Domestic and foreign scholars have conducted a lot of research on the relationship between urban transportation and land use, and established related models. Domestic research mostly focuses on static qualitative analysis [2], such as land intensive use theory, urban scale theory, traffic capacity theory, etc. [3, 4, 5, 6]. These theories are mostly based on static analysis, and cannot fully clarify the interaction mechanism of land use and transportation systems in the interactive process. On the integration of urban transport and land use, representative methods include: methods based on Lowry theoretical model, methods based on mathematical programming, methods based on spatial input-output, methods based on economic theories, etc. [7, 8, 9, 10, 11, 12]. In these methods, the establishment of models not only requires complex functions and details, but also requires a large amount of basic data as analysis support, which leads to defects such as high cost, long cycle, lack of focus, and biases forecasts when analyzing the relationship between the two. In the course of practice, the model establishment also mostly relies on the hypothesis of urban single-center development, which is a certain discrepancy with actual development [13]. Therefore, many scholars are actively exploring the research of model simplification and model validity. They can not only quantitatively study the interaction between the elements in the interaction between the two, but also dynamically feedback and predict the synergistic relationship between transportation planning and land use.

In view of this situation, this article first proposes a combination of dynamic and static transportation planning and land use collaborative optimization system. The initial key stage of this system is to simplify the model and extract key information, in other words, the determination of system units. The determined system unit is closely related to each link of the system. Thus, this paper not only introduces this method of collaborative optimization of transportation planning and land use, but also focuses on the link of system unit decision-making. In view of DEMATEL can well determine the centrality, cause degree and affected degree of element elements, this paper will adopt DEMATEL method to determine the influence element.

2. Overview of the collaborative optimization system for transportation planning and land use

This research comprehensively uses DEMATEL model, system dynamics model, structural equation, DEA evaluation and BP neural network technology to construct a complete transportation planning and land use collaborative optimization system.

First, at the macro level, combined with DEMATEL model, the impact units of transportation planning and land use are sorted out to clarify the interaction mechanism of the two, and the key impact units are extracted to form a macro interaction system. Then the extracted influence units are further micro-decomposed. Combined with structural equation and system dynamics, the dynamic and static relationship between transportation planning and land use is studied at the micro level. Combined with structural equation and system dynamics, the dynamic and static relationship between transportation planning and land use is studied at the micro level. On this basis, the DEA evaluation method is used to evaluate the degree of synergy between transportation planning and land use. According to different evaluation results, neural network technology is used to forecast and adjust transportation planning and land use. Through the above process, the collaborative optimization of transportation planning and land use is finally formed. The specific technical framework is shown in Fig. 1.

Figure 1.

Technical framework.

3. The macro-interaction mechanism between urban transportation planning and land use

3.1 System unit determination

The interaction between transportation planning and land use is a complex binary system. The entire system moves, develops, and changes under non-equilibrium conditions, and each influencing factor often has a multi-cause and multi-cause relationship. Due to the constant changes of population, economy, environment, climate, etc., the coupling correlation between various elements has become more and more ambiguous and complicated [14]. Using the semantic network method, we can find that there are various coupling relationships between transportation planning and land use. By analyzing the coupling and causality of urban elements, a large amount of associated data can be mined [15].

In the semantic network diagram shown in Fig. 2, the change of any one element will affect other elements. In the process of interaction between transportation system and land use, there are often more complex network relationships between elements than semantic network diagrams. For example: In the planning process, starting from land development and utilization, if the land utilization rate becomes higher and higher, it will generally lead to an increase in travel generation. This also puts forward higher requirements on transportation facilities. With the improvement of transportation facilities, accessibility of traffic has also improved, while attracting people to continue to invest in human and material resources in the area, forming a process of positive feedback. However, in the process of urban development, it is difficult to increase traffic capacity after certain urban transportation facilities have developed to a certain extent. What’s more, as the pace of urbanization in my country accelerates, in order to balance the needs of suburban expansion and highway construction, many places plan a more livable, sustainable development, and convenient transportation community land model in the region. Therefore, in the collaborative process of transportation planning and land use, more potential variables are involved, which increases the complexity of the model [16].

Figure 2.

Semantic network diagram of the relationship between urban land use and transportation facilities.

Usually in the process of model building, the characteristics of different factors affecting the entire model are often ignored, blindly pursuing to cover many fields or blindly pursuing large-scale data survey statistics, resulting in a huge consumption of manpower and funds, and some survey statistics. The delay will affect the effectiveness and accuracy of the entire model [17].

Therefore, it is necessary to rationally simplify the model, select valuable system units, and carry out analysis and adjustments. It can not only sort out the hierarchical and nested relations between the influence units, but also provide a scientific basis for analyzing the influence relations between the system units. This paper intends to use the DEMATEL model to construct a causality diagram and an incidence matrix for the above system units, so as to identify the central influence unit and the causal influence unit, clarify the macro-interaction mechanism between urban traffic and land use, and carry out the extraction of the key influence units. The subsystem of micro elements is constructed to make preliminary preparation for the analysis of static and dynamic integration model. This part is the focus of this paper.

3.2 DEMATEL’s system unit decision

From the above discussion, we can see that an effective simplified model is the basis for the study of collaborative optimization of land use and transportation planning. After that, each link is related to the determination of system units. Therefore, the determination of the system unit is the key link of this method. The following will focus on the DEMATEL-based system unit decision-making process.

Basic steps:

Step 1: Build a direct impact matrix.

On the basis of extracting all system units, the system units are compared in pairs. It should be noted that the causality of each system unit must be considered in the process of pairwise comparison, so the constructed matrix is not a symmetric matrix. The determination of the strength of the relationship can indirectly reflect the degree of mutual influence between the units through expert scoring and questionnaires, etc., using a 5-level scale (0 means no influence, 1 means minor influence, 2 means general influence, 3 Is a large impact, 4 is a very large impact).

Constructing incidence matrix by scale $A=({a_{ij}})_{m\times n}$ , where $a_{ij}$ represents the degree of influence of the $i$ system unit on the $j$ system unit.

Step 2: Return to “1” the incidence matrix $X$ , $X=m.A$

$\displaystyle m=\frac{1}{\mathop{\max}\limits_{1\leqslant i\leqslant n}\mathop% {\sum}\nolimits_{j=1}^{n}a_{ij}},i,j=1,2,3,\ldots,n$ (1)

Then

$\displaystyle X=({x_{ij}})_{m\times n}=\frac{1}{\mathop{\max}\limits_{1% \leqslant i\leqslant n}\mathop{\sum}\nolimits_{j=1}^{n}a_{ij}}.A$ (2)

Step 3: Solve the comprehensive influence matrix $T$ , $T=X({I-X})^{-1}$ , $({I-X})^{-1}$ is the inverse matrix of $I-X$ , $I$ is the identity matrix.

Step 4: Determine the centrality and cause of each system unit.

$\displaystyle T=({t_{ij}})_{m\times n}$ (3) $\displaystyle D=({t_{i}})_{n\times 1}=\left({\mathop{\sum}\limits_{j=1}^{n}t_{% ij}}\right)_{n\times 1}$ (4) $\displaystyle C=({t_{J}})_{1\times n}=\left({\mathop{\sum}\limits_{i=1}^{n}t_{% ij}}\right)_{1\times n}$ (5)

$D_{i}+C_{i}$ as the centrality of the system unit, the centrality represents the position of the system unit in the evaluation index system and the size of its function. The larger the value, the higher the importance in the entire system. $D_{i}-C_{i}$ is the cause degree of the system unit. If the cause degree is greater than 0, it indicates that the unit has a great influence on other influencing units and is called the cause influencing unit. Otherwise, it is called result influence unit.

3.3 Influencing unit identification

(1) The influence unit in the interactive process of urban land use and transportation planning is composed of a series of macro-elements with internal connections, which can reflect the mutual influence in the interactive process from multiple angles, and then scientifically extract the key influencing in the interactive process. Finally, we will analyze and adjust them.

Combining urban overall planning, regulatory detailed planning, transportation planning, related the domestic and foreign research, and integrating the five categories of land use, transportation, population, economy, and environment, a macro-element index system that affects the coordinated development of urban land use and transportation planning can be obtained. As shown in Table 1.

Table 1
Classification of macroscopic impact units

Land use	Traffic	Population	Economic
Land use intensity (M1)	Traffic demand (M7)	Demographic changes (M12)	National economic situation (M13)
Land price (M2)	Traffic supply (M8)
Land use type (M3)	Traffic accessibility (M9)
Geology (M4)	Number of trucks (M10)
building height (M5)	External traffic (M11)
Architectural style (M6)	1.833

(2) Incidence matrix

In order to prevent the development of different cities from affecting the accuracy of the model’s results, this paper selects 7 cities: Beijing, Shanghai, Xiamen, Shenyang, Dalian, Hainan, and Harbin. Through experts scoring the relationship between the impact units of different cities and issuing questionnaires, the correlation matrix between the influence units is constructed. As shown in Fig. 3.

3.4 Calculation results

Through the influence matrix, we can see that Maxvar $=$ 36. Then further and further regulate the “1” processing and standardize the inverse matrix of the direct of the direct relationship matrix. The final settlement result is shown in Table 2 and Fig. 4. According to $D_{i}+C_{i}$ , in the process of interaction between transportation planning and land use, with “5” as the demarcation point, there are 8 impact units with greater centrality, which are land price, population, transportation supply, transportation accessibility, land use intensity, national economy, transportation demand, and land use types. However, factors such as building height, external traffic, and nitrogen oxide emissions have a weaker impact on the overall structure. This is because these influencing units are used as intermediate links in the interactive process. The data is mainly formulated based on the influencing units with strong centrality.

Table 2
Influence unit centrality and cause degree

$M_{14\times 4}$	$D_{i}$	$C_{i}$	$D_{i}+C_{i}$	$D_{i}-C_{i}$
M1	4.376	4.112	8.488	0.264
M2	4.257	4.624	8.881	$-$ 0.367
M3	3.749	3.6	7.349	0.149
M4	2.151	2.657	4.808	$-$ 0.506
M5	2.117	2.739	4.856	$-$ 0.622
M6	1.833	1.743	3.576	0.09
M7	4.199	4.272	8.471	$-$ 0.073
M8	4.431	4.229	8.66	0.202
M9	4.402	4.2	8.602	0.202
M10	2.193	2.058	4.251	0.135
M11	2.301	2.397	4.698	$-$ 0.096
M12	4.506	4.358	8.864	0.148
M13	4.571	3.908	8.479	0.663
M14	2.236	2.425	4.661	$-$ 0.189

Figure 3.

Influencing unit incidence matrix.

Figure 4.

Causality diagram of influence unit.

According to $D_{i}-C_{i}$ , in the process of interaction between transportation planning and land use, the causal influence unit is 8 factors such as land use type, land use intensity, transportation supply, and national economy. The remained 6 units are result-type influence units. It shows that these 6 result-type influence units are the influence results of 8 cause-type influence units.

In summary, the key macro-influencing units that affect the coordination and interaction of transportation planning and land use are: land price, population, transportation supply, transportation accessibility, land use intensity, national economy, transportation demand, and land use type. Any change in one element will have an impact on other elements. In the planning process, starting from land development and utilization, if the intensity of land use and land types are adjusted, it will inevitably lead to changes in travel generation. This also puts forward corresponding requirements on public transportation, transportation and other transportation infrastructure. As the transportation supply changes, the accessibility of transportation also changes accordingly, and at the same time affects people’s human attraction and material investment in the area. In order to make the above-mentioned causality tend to be positive feedback or stable, it is necessary to decompose these established macro-influencing units, and then clarify the micro-interaction mechanism and mutual influence relationship between the decomposition elements.

4. The micro-interaction mechanism between urban transportation planning and land use

4.1 Urban transportation planning and micro-static analysis of land use

Using SEM structural equations, we construct a microscopic structural model and measurement model of the interaction between transportation planning and land use. First of all, the system unit extracted from the DEMATEL model is used as the latent variable in the model, and the structural model is constructed through the causality of the exploratory assumption. Then, in a progressive way, these system units are microscopically decomposed and decomposed into measurable system elements as observation variables. Here we construct measurement models corresponding to observed variables and latent variables. Finally, the data of latent variables are obtained through measurement tools such as questionnaires or scales, and imported into the model. The covariance matrix between variables is used to test the significance of the observed variables under non-standardization, the importance of the observed variables under standardization, the reliability and fit between the overall variables and the actual data [18]. The variables or causal relationships that do not meet the conditions are adjusted to construct a hypothetical model suitable for sample data. So that we will clarify the micro-static interaction mechanism between transportation planning and land use, and obtain the dynamic equation between the variables [19].

For example, the supply-demand matching relationship between land use and transportation supply is simply simulated to establish an SEM model. Then we decompose land use into four elements: construction land, land use pattern, land use compatibility, and floor area ratio. Changes in each factor will trigger changes in land value and travel value. If the traffic demand increases, it will put forward corresponding requirements on the traffic supply. The expansion of traffic capacity and the improvement of traffic facilities can improve the accessibility of the land and change the travel mode of travelers, so as to achieve a good matching relationship in Fig. 5.

Figure 5.

Schematic diagram of static analysis.

4.2 Urban transportation planning and micro dynamic analysis of land use

Proposing an optimal urban transportation system model to achieve an interactive balance between transportation and land use has always been a hot spot for scholars [20]. Using the method of system dynamics, we can further study the dynamic correlation of the elements that affect transportation planning and land use in the interactive process. First, we determine the system boundary and system variables, and convert the key influencing units identified by the DEMATEL model into dynamic system boundaries. The system unit and its subsystems are used as system variables, and a causal relationship diagram is drawn according to the causal relationship between the variables.

As shown in Fig. 6, the increase in residential land has led to an increase in the total construction land. Due to the increase in urban development intensity, the floor area ratio has increased and the travel ratio has risen, which caused greater traffic demand. It requires more traffic supply and increases the traffic accessibility of residential land. Then, inducing new residential land. In the same way, the increase of residential land will lead to an increase in the total construction land, an increase in the plot ratio and an increase in the value of the land. The increase in land price promotes the construction of roads improves the supply of traffic, and less traffic jams have stimulated the development of residential land. Other land types also have similar feedback effects. Through this closed feedback loop, the dynamic interaction mechanism between traffic planning and land use is simulated. Then through the commensurability transformation of the semantic network model, the dynamic model can be further obtained [21].

Next, according to the dynamic equation obtained by SEM, the system flow diagram is described quantitatively. We use Vensim software to establish a dynamic model of transportation planning and land use system through statistical regression analysis, clarify the feedback mechanism between variables, perform simulation output and error analysis on the results, and clarify arbitrary level variables and their interactive relationships with causal constants, which used for subsequent collaborative optimization evaluation and policy analysis. For example, we can control the plot ratio and monitor changes in land prices and traffic demand by adjusting the growth of building area, or monitor the degree of changes in traffic congestion by controlling the number of new rail transit vehicles, finally determine the trend of changes in the area of building land.

Figure 6.

Schematic diagram of dynamic analysis.

4.3 Evaluation of interaction between urban transportation planning and land use

About the complex interactive relationship between urban traffic and land use, in the process of evaluating the coordinated relationship between the two, for the purpose of avoiding the subjective arbitrariness of artificially determining weights in traditional evaluation methods, non-parametric statistical analysis methods are proposed to be used for existing planning to evaluate, which is the DEA method (data envelopment analysis). First, divide the pre-evaluation area into several groups, which are used as evaluation units to form multiple DMUs [22]. Then, we decompose the key influencing units and regard the urban transportation system and the land use system as input and output systems that are mutually input and output. For example, land use is an input variable, including population density, construction land, employment-to-living ratio, etc.; traffic status as output variable, which encompasses the average travel distance, public road traffic, rail traffic supply, traffic demand, etc. What’s more, through the analysis of the input and output ratio, the effective production frontier is determined [23]. Finally, through the distance between each DMU and the production frontier, it is determined whether it is effective, and the efficiency value is obtained. In other words, it is determined the degree of coordination between traffic and land use in each group.

4.4 Cooperative optimization forecast and adjustment of urban traffic planning and land use

The BP neural network model is used to make use of its nonlinear characteristics to predict the overall satisfaction degree of traffic planning and land use in the planning process [24]. Above all, we select the quantifiable variables in the static analysis and the relevant indicators that may affect the efficiency in the DEA model as the n nodes of the input layer and the hidden layer, and construct the node matrix [25]. Then, we take the effective group of DEA as the learning and training sample, input the data, make the efficiency of DEA as the expected value of the output, compare the expected value and the output result, adjust the node parameters of each layer through multiple forward propagation and reverse convergence, until the error between the output result and the expected value is the smallest. So, a more stable neural network structure is formed. At last, we transmit the non-valid cluster data in DEA to confirm the structural stability. The stable neural network model can scientifically predict the satisfaction degree of the collaborative optimization of transportation planning and land use.

In the optimization and adjustment process of the control detailed planning, referring to the BP output result and according to whether the coordination optimization degree value is satisfied or not, we will compare the influence unit of the model and feed back to control detailed planning, and select control elements with optimization potential in the previous dynamic model for combination optimization. Through the constructed system dynamics model and comparing the stability of the elements, we optimize and adjust the model, and indirectly update the BP output results. If the BP value of the optimized control scale is also satisfactory, the staged collaborative optimization process is completed. At the same time, the optimized control element model can be used in the next step of planning and design. If there are no unacceptable conditions in the process of control planning, it can be considered that the control index system and its traffic planning indicators have reached a comprehensive level and satisfaction level.

5. Conclusion

Through the commensurability of the semantic network model, we can realize the association and transformation between the system unit model and the system dynamics model, structural equation model, and evaluation and prediction models. In addition, we can clarify the relevance of influencing factors, and support the construction of evaluation and simulation models. Through research, we can know that DEMATEL’s system unit decision-making process is relatively objective and scientific. The set of system units obtained in this way can effectively analyze the key influencing factors, eliminate a large number of weakly related factors, and avoid interference from irrelevant factors. Determining the system units of DEMATEL can increase the efficiency of collaborative optimization of transportation planning and land use and save computing costs. It is the key link and foundation of the collaborative optimization of transportation planning and land use.

To sum up, in the process of studying the collaborative optimization of land use and transportation planning, the centrality of the influence unit and the causal influence unit are regarded as the key influence units at the macro level, which are the next important influence units. It provides an accurate research direction for the micro-decomposition and research analysis of influencing units.

Footnotes

Acknowledgments

This research is supported by “the Fundamental Research Funds for the Central Universities” DUT19RC (3) 044 and “National Natural Science Foundation of China” (No. 51278158).

References

Tian

Niu

. Evaluation of the coordination relationship between urban land use and urban transportation: Taking the main urban area of Chongqing as an example. J Southwest Univ (Nat Sci Ed). 2019; 41(7): 96-104.

Yang

. Research on the basic theories and methods of the relationship between urban traffic and land use [dissertation]. [Beijing]: Beijing Jiaotong University; 2007.

Pei

Zhang

. Research on visual recognition effect of different areas of guide sign layout based on visual recognition characteristics. Forest Eng. 2020; 36(4): 109-115.

Huang

. The adaptive transformation of transportation planning under the new territorial and spatial planning system: Taking Xiamen as an example. Urban Plann. 2019; 43(7): 21-33.

Hou

. Research on collaborative optimization of urban land use intensity and traffic capacity in hierarchical control planning [dissertation]. [Chang’an]: Chang’an University; 2015.

. Evaluation and optimization path of coupling coordination between land use efficiency and economic development level [dissertation]. [Harbin]: Harbin Normal University, 2019.

Lowry

. A model of metropolis, rand corporation. Santa Monica, California: R-and Corporation, 1964.

Steger

. The Pittsburgh urban renewal simulation model. J Am Inst Plann. 1965; 31(2): 144-150.

Mackett

. LILT and MEPLAN: A comparative analysis of land-use and transport policies for Leeds. Transp Rev. 1991; 11: 130-131.

10.

Mills

. Mathematical models for urban planning: Urban and social economics in market and planned economics [dissertation]. [New York]: Praeger; 1974.

11.

Ellickson

. An alternative test of the hedonic theory of housing markets. J Urban Eco. 1981; 9(1): 56-79.

12.

Justin

Roger

. A bi-level model of the relationship between transport and residential location. Transp Res Part B: Methodol. 2006; 40(2): 123-146.

13.

Zhao

. Foreign urban land use-transportation system integration model. Econ Geogr. 2000; (6): 79-83+128.

14.

Leibowicz

. Urban land use and transportation planning for climate change mitigation: A theoretical framework. Eur J Oper Res. 2020; 284(2): 604-616.

15.

Dong

Zou

. Urban design logical scheming based on semantic network. Inf Technol J. 2013; (12): 5725-5730.

16.

Sabouri

Dillon

Proffitt

Townsend

Ewing

. State-of-the-practice in connecting and coordinating transportation and land use planning in the USA. Transp Res Rec. 2019; 2673(9): 240-253.

17.

Zhang

Yang

. Research on the interactive relationship model of land use and transportation. Highw Transp Sci Technol. 2009; 26(6): 114-118.

18.

Cheng

. Characteristics and application of structural equation model. Stat Decis. 2006; (10): 22-25.

19.

Xie

. Research on the system dynamics model of the interaction between urban land use and transportation [dissertation]. [Huazhong]: Huazhong University of Science and Technology, 2017.

20.

Sun

Gao

Ding

. Optimal urban expressway system in a transportation and land use interaction equilibrium framework. Transportmetrica A: Transp Sci. 2019; 15(2): 1247-1277.

21.

Dong

. Commensuration operator of urban science based on semantic network – A case study of urban design programming. Q J Indian Pulp Pap Tech Assoc. 2018; (30): 356-361.

22.

Wei

. Data envelopment analysis (DEA). Sci Bull. 2000; (17): 1793-1808.

23.

Yuan

Yan

Zhang

Wang

. DEA model of land use and traffic coordination of planning control unit with preference constraints. J Traffic Transp Eng. 2017; 17(6): 86-96.

24.

Shen

Wang

Gao

Qin

Yao

. Determination of the number of hidden layer units of BP neural network. J Tianjin Univ Technol. 2008; (5): 13-15.

25.

Huang

. BP neural network algorithm improvement and application research [dissertation]. [Chongqing]: Chongqing Normal University; 2008.

Collaborative optimization method of transportation planning and land use and its system unit decision making

Abstract

Keywords

1. Introduction

2. Overview of the collaborative optimization system for transportation planning and land use

3.1 System unit determination

Table 1 Classification of macroscopic impact units

Table 2 Influence unit centrality and cause degree

4.1 Urban transportation planning and micro-static analysis of land use

4.4 Cooperative optimization forecast and adjustment of urban traffic planning and land use

5. Conclusion

Footnotes

Acknowledgments

References

Table 1
Classification of macroscopic impact units

Table 2
Influence unit centrality and cause degree