Research on trans-region integrated traffic emergency dispatching technology based on multi-agent

Abstract

With the increasingly close links between provinces and municipalities, the complexity and social connectivity of emergencies have gradually increased. How to realize the rapid and scientific dispatch of emergency resources based on the comprehensive transportation network has become a realistic need. However, due to barriers in management and communication between different administrative regions and different modes of transportation, emergency resource scheduling can only be carried out in the region, and the utilization of integrated traffic channels is insufficient to form a good multimodal transport system. The multi-agent system has strong self-organization ability, learning ability and reasoning ability, which solves problems in the fields of dynamic decision-making and micro-simulation and provides ideas for solving the problem of comprehensive traffic emergency dispatching across regions. In this paper, the multi-agent system is introduced into the research of the task assignment problem of trans-regional comprehensive traffic emergency materials dispatching, based on the traditional bidding rules, the emergency dispatch task assignment rules are proposed and multi-agent trans-regional comprehensive traffic emergency dispatching model based on improved bidding rules is established. The simulation results show that the proposed method can break the regional barrier and form a multimodal transport scheme for emergency materials from the reserve point to the demand point under the minimum generalized time cost, providing decision support for emergency dispatch.

Keywords

Multi-agent trans-region emergency dispatch bidding

1 Introduction

In the face of emergencies such as earthquakes, epidemics, terrorist attacks, etc., the relevant departments of the world are dealing with the disaster situation according to the pre-established emergency plan [1 –3]. Emergency material dispatching, as the core of disaster relief missions, has always been the focus of research by experts and scholars. The emergency materials dispatching work mainly includes analyzing the demand for emergency materials, assigning rescue teams, and planning transportation routes [4, 5]. The essence of the problem is decision-making and optimization.

At present, for the study of emergency material dispatching issues, scholars mostly take the perspective of “Central Controller” as the starting point, aiming at accelerating emergency response time and reducing disaster losses, establishing scheduling models and using relevant algorithms to solve [6 –9]. In addition, some studies have also considered the cost of resource scheduling, as well as the emergency demand point and the state change of the integrated road network during the scheduling process [10 –12] in differrent cases. These research results provide a lot of inspiration for the paper in terms of models, algorithms and simulations. The deeper research can be carried out on the following aspects:

(1) There are few studies on trans-regional dispatch of emergency materials. Due to the different scope of responsibility and authorization of the government and functional departments, there is a rigid segmentation of emergency management between regions [13], which also leads to related research mainly for the same region [7], as well as within the city [6]. However, with the development of social economy, the links between provinces and municipalities have become increasingly close, and the complexity and social connection of emergencies have gradually increased. When the scope and destructiveness of emergencies exceed the ability of a single regional response, a scientific and efficient trans-regional emergency linkage mechanism is the basic guarantee for the rapid handling of emergencies.

(2) Only a single mode of transportation, such as highway [7], railway [3,17 , 22], or helicopter [9], is not fully considered to utilize the integrated traffic corridor to form an emergency intermodal system. As a connecting channel between regions, the comprehensive transportation network plays an important role in ensuring the timely transportation of relief materials and the timely placement of rescue workers. It is necessary to make full use of the advantages of various modes of transportation to realize emergency resource dispatching in multiple modes of transportation.

(3) The increase in the complexity of emergencies will lead to troubles of large-scale multi-source and multi-sink (“multiple rescue points – multiple demand points”) problems, resulting in a sharp increase in system computing.

The multi-agent system has the characteristics of distribution, autonomy and coordination [14 –16], which provides new ideas for solving large-scale trans-regional integrated traffic emergency dispatch problems. In Section 2, we establish an emergency material allocation model based on the bidding rules by analyzing the characteristics of emergency materials dispatching. In Section 3, we describe the rules of the model and the specific behavior of each type of agent, with a focus on how to achieve inter-regional multimodal transport. In Section 4, we study the algorithm and gives examples.

2 Emergency resources scheduling model based on bidding rules

2.1 Multi-agent bidding rule model

In a multi-agent system, each agent learns to improve its own strategy [19, 20] by interacting with the environment to obtain reward values to obtain an optimal strategy in the environment. A multi-agent system can be expressed as (n, A₁, A₂, ⋯ A_n, R₁, R₂, ⋯ R_n) where n means the number of agents, A_i means the action set of the ith agent, and R_i explains A₁ × ⋯ × A_n → R, which represents the reward function of the ith agent. It can be seen from the reward function that the reward obtained by each agent is related to the linkage action [18] of the multi-agent system. The joint action space is named as A₁ × ⋯ × A_n. V_i (Π₁, ⋯ , Π_i ⋯ Π_n) is defined as the reward of the agent i under the coupling strategy (Π₁, ⋯ , Π_i ⋯ Π_n), which is also called the value function [21].

Bidding is a kind of transaction method adopted for contracting and contracting service projects under the conditions of market economy. The bidding strategy is a collection of series of plans to achieve the objectives, including the bidding proposal, bid evaluation method and winning bidding method. Usually, a large project will be split into several tenders and tendered separately. A unit can bid for multiple packages, and multiple targets at the same time, and can also be subcontracted. Multiple units can form a joint venture to bid, and jointly take charge of the project after winning the bid.

In a multi-agent system, the bidding rule model can be described by the following formula (1): $Bid = 〈 T, A, B, F 〉$ (1)

In formula (1), T means a bidding strategy set, A means a bidding agent set, and B means a bidding agent set. The bidding process is a typical dynamic game process. The tenderer mainly focuses on the qualifications and service quality of the bidders. At the same time, considering the bidder’s quotation, the bidder pays more attention to the income and risk, and takes into account the reputation of the company. Therefore, it is necessary to introduce the system comprehensive benefit index F, to measure the status of the multi-agent bidding system.

2.2 Multi-agent emergency scheduling model

Appropriate improvement of the bidding rules model can be used to allocate emergency dispatching tasks. The basic model is as follows: $Game = 〈 M, S, D, C, T, E 〉$ (2)

In formula (2):

M means a set of scheduling task objects that need to be completed, M ={ M₁, M₂, ⋯ , M_N };

D means a collection of demand point agents, D ={ D₁, D₂, ⋯ , D_N }, corresponding to the tenderer, responsible for receiving emergency supplies, and there are n demand points;

S means a collection of rescue points, S ={ S₁, S₂, ⋯ , S_N }, corresponding to the bidder, responsible for providing emergency supplies, and there are N rescue points;

C means the set of expenses that the rescue point agent needs to pay to complete the rescue mission, $C = {C_{S_{1}}^{M_{1}}, C_{S_{1}}^{M_{2}}, \dots C_{S_{1}}^{M_{N}}, \dots, C_{S_{N}}^{M_{N}}}$ , which represents that S_i means the cost of completing the task S_i, M_j, i ∈ (0, n] , j ∈ (0, N];

T means the set of time that the rescue point agent needs to pay to complete the rescue mission, $T = {T_{S_{1}}^{M_{1}}, T_{S_{1}}^{M_{2}}, \dots T_{S_{1}}^{M_{N}}, \dots, T_{S_{N}}^{M_{N}}}$ , which represents that S_i means the time taken to complete the task T_j, i ∈ (0, n] , j ∈ (0, N];

E means the set of benefits generated by the demand point after receiving emergency supplies, $E = {E_{D_{1}}^{M_{1}}, E_{D_{1}}^{M_{2}}, \dots E_{D_{1}}^{M_{N}}, \dots, E_{D_{N}}^{M_{N}}}$ , which represents the benefit of D_i receiving disasters from the transportation of the mission M_j, i ∈ (0, n] , j ∈ (0, N].

It can be seen from the above model that C and T are both cost type indicators, so the linear weighting of rescue time and cost can be defined as “generalized time cost” to describe the comprehensive cost generated by the scheduling process. Therefore, the generalized time cost function defining the rescue point agent is as follows: $G = \frac{- 1}{λ} l n \sum_{i, j} e x p [- λ (α C + (1 - α) T)]$ (3)

$G = {G_{S_{1}}^{M_{1}}, G_{S_{1}}^{M_{2}}, \dots G_{S_{1}}^{M_{N}}, \dots, G_{S_{N}}^{M_{N}}}$ , indicating that S_i means the generalized time cost to be paid for completing the task M_j, i ∈ (0, n] , j ∈ (0, N].

Therefore, the optimal task allocation scheme in a multi-agent system should satisfy: $max \sum_{j = 0}^{N} \sum_{i = 0}^{n} (E_{j} - G_{D_{i}}^{M_{i}})$ (4)

The above research shows that for the characteristics of trans-regional comprehensive traffic emergency dispatching, appropriate improvement of multi-agent bidding rules can fully mobilize the autonomy and enthusiasm of each agent, strengthen the communication and coordination between demand points and rescue points, and each rescue points to achieve optimal assignment of tasks.

3 Emergency resource dispatch task assignment

3.1 Emergency dispatch rules

The emergency dispatching process is shown in Fig. 2. The demand point agent decomposes the emergency rescue task into multiple target segments (sub-tasks) and then invites the target. The task of each target segment can be completed by a rescue point agent or multiple rescue points. Each winning bidding agent can also moderately “subcontract” the task to other rescue point agents according to the actual situation and work together.

Fig. 1

Schematic diagram of multi-agent emergency dispatching model based on bidding rules.

Fig. 2

Trans-regional emergency dispatch process based on multi-agent.

3.2 Demand point agent behavior

3.2.1 Demand point agent workflow

The demand point agent workflow is shown in Fig. 3. After the disaster occurred, the emergency rescue command office will be set up in the disaster-stricken region to generate the demand point agent, count the disaster situation and emergency resources demand information, such as time, place, disaster type, disaster description, resource demand type, resource demand quantity and so on, set the effective time limit Q for emergency rescue, divide the list of emergency resources requirements into a number of “standard sections”, and recruit and invite bids for all rescue points according to the emergency dispatch rules.

Fig. 3

Demand point agent workflow.

After the bidding information is issued, the demand point agent receives a number of bidding applications within the effective time limit Q, and the demand point agent selects the “winning bidder” according to the utility function experience value of each rescue point agent, and issues a bid winning notice, and begin to receive emergency supplies for the transportation of the rescue point.

After the bidding information is issued, the demand point agent receives a number of bidding applications within the effective time limit, and selects the winning bidder by evaluating the qualification of each salvage point agent (ie, judging the value of the utility function value), and issues a bid winning notice. The demand point agent begins to receive emergency supplies and continuously adjusts the demand information according to the actual situation until the disaster situation is completely eliminated. After the disaster situation is eliminated, the demand point multi-agent corresponds to the whole process of the emergency disaster, and the utility function value of each rescue point agent participating in the disaster relief operation is given.

3.2.2 Demand point agent information description

As the representative of the Disaster Relief Command Office and the Emergency Material Distributing Center of the disaster-stricken region, the demand point agent collects statistics on the disaster situation and emergency materials demand information based on the comprehensive traffic road network, and the disaster relief task is decomposed into several “standard sections” for bidding and valid time limit is set. The information is described as follows:

<Agent_Tenderee>::=<Basic Information> <Disaster Description> <Requirement Information> <Bidding Rules> {

<Basic information>::=(number, address, railway accessibility, air accessibility, road accessibility, warehouse maximum reserve capacity)

<disaster description>::=(time,<set of disaster points>) {<set of disaster points>::=(number, address, type of disaster, casualty information, responsible unit)}

<Requirement Information>::=(<Responsive Point Collection>, Demand Material Type, Demand Material Quantity)

<Tendering Rules>::=(Principal Division Principle, Inviting Bidding Objects, Open Bidding Objects, Historical Bidding Information, Valid Time Limit)

}

Based on the above information, the demand point agent creates a bidding document as shown in Fig. 4. Among them,<segment task description>::=(required material type, quantity of demanded goods).

Fig. 4

Demand point agent bidding file format.

(1) Task decomposition

Under the emergency, the material distribution center has a large demand for the types and quantities of emergency materials, and needs to be transferred from multiple material reserve points. Based on the types of emergency materials, the task is decomposed into several “standard sections” for bidding.

(2) Public bidding and invitation bidding

The rescue point agent in the region is in principle obligated and necessary to participate in emergency dispatch, so the demand single agent adopts the “call” mode. When the rescue point agent in the same region cannot meet the emergency dispatching demand for various reasons, the demand single agent body faces the open bidding for the rescue point agent outside the region.

(3) Utility function calculation and estimation

The utility function value is an indicator to measure the quality of the bidder’s completion of the task. The calculation of the utility function value is as shown in equation (5). In the initial state of the system, the utility value of each rescue point utility defaults to 0. The utility function value information calculated by each demand point agent is public. The winning point of the successful bidding agent exits for its own reasons, and the utility function will be affected. $E_{D_{i}}^{M_{j}} = w_{1} c_{1} q (M_{j}) + w_{2} c_{2} RT (M_{j}) - ɛ$ (5)

In formula (5), q (M_j) represents the amount of resources transported for the task M_j, c₁ represents the benefit of the relief of the disaster caused by the unit resources, R represents the probability of further expansion to the potential risk, T(M_j) represents the fastest emergency response time for the task M_j, c₂ represents the unit loss under the condition that the demand point does not receive emergency resources, w₁ and w₂ are the weight coefficient, and ɛ is the residual.

In the process of bidding, the bidders often receive applications from bidders who have never established a cooperative relationship, so it is necessary to evaluate the qualifications of the bidders. The evaluation is mainly based on the relevant historical data (i.e., the empirical value of the utility function of the emergency dispatching process), and the value of the utility function generated by the bid is estimated by the formula (6). The utility function value estimation method is also applicable to the bidder’s decision behavior. $E_{D_{i}^{'}}^{M_{j}^{'}} = σ E_{D_{i}}^{M_{j}^{'}}$ (6) $σ = \frac{\sum_{i} \sum_{j} (\frac{d^{*}}{DDT})}{n \sqrt{\frac{(n - 1) {LWT}^{2}}{{AWT}^{2}}}}$ (7)

In equation (6), the execution task $M_{j}^{'}$ is expected to produce a utility value $E_{D_{i}^{'}}^{M_{j}^{'}}$ obtained by multiplying the historical actual utility function value $E_{D_{i}}^{M_{j}^{'}}$ by an empirical operator σ. The empirical operator σ calculation method is shown by the formula (7), where n is the number of performance samples selected by the tenderer for the research bidder qualification, and d^* is the road network distance between the demand point agent and the bidding rescue point agent. DDT is the total transit time for the task to perform, LWT is the fastest response time, AWT is the average response time.

3.3 Rescue point agent behaviour

3.3.1 Rescue point agent workflow

The rescue point agent workflow is shown in Fig. 5. Each emergency material reserve point distributed nationwide corresponds to a rescue point agent. Each agent collects the bidding information within the system, judges whether to bid according to the emergency dispatching rules, determines the bidding and then calculates and calculates the transport path based on the integrated traffic dynamic road network, and obtains the transportation plan and sends it to the demand point agent.

Fig. 5

Rescue point agent workflow.

After receiving the bidding notice of the demand point agent, the transportation task can be completed independently, or part of the task can be subcontracted to other rescue point agents. Once the subcontracting is performed, the agent role is transformed into the demand point agent. If the task cannot be completed due to various reasons, the rescue point agent needs to report to the demand point agent in time for further processing. $E_{D_{i}}^{M_{j}} - G_{S_{i}}^{M_{j}} \geq 0$ (8)

$\begin{matrix} min [α (\sum_{j} \sum_{k} q_{j}^{m} c_{ij}^{k} d_{ij}^{k} x_{ijk} + \sum_{k} \sum_{l} q_{j}^{m} z c_{i}^{kl} y_{ikl}) \\ + (1 - α) \sum_{j} \sum_{k} q_{j}^{m} \frac{d_{ij}^{k}}{v_{ij}^{k}} x_{ijk}] \end{matrix}$ (9)

Equation (8) indicates that the benefit of the rescue point S_i execution task M_j for the disaster point D_i is greater than the generalized time cost paid by the rescue point D_i;

The solution result of equation (9) is the optimal transportation plan for the emergency resources of the disaster point D_i $q_{j}^{m}$ is the number of resources in the m-th point of the rescue point intelligent demanding point j transportation; $c_{ij}^{k}$ is the transportation cost of the unit distance of the kth transportation vehicle for the emergency relief supplies from node i to node j; ${zc}_{i}^{kl}$ is that the emergency rescue resource is the unit transfer fee from the kth transportation mode to the lth transportation mode at the node i; $t_{i}^{kl}$ is the time when the goods are transferred from the kth transportation vehicle to the first transportation vehicle at the node i; $d_{ij}^{k}$ is the emergency resources selected from the node i to the node j by the transport distance i of the kth transport mode; $v_{ij}^{k}$ is the transport speed of the kth transport vehicle selected from the node i to the node j.

3.3.2 Information description of rescue point agent

The rescue point agent needs real-time monitoring and management of the storage materials, and plans the emergency material dispatching and transportation plan based on the comprehensive traffic road network information and the accessibility of each transportation mode. The specific information is described as follows:

<Agent_Tenderer>::=<Basic Information> <Material Reserve Information> <Integrated Traffic Network Information> <Bid Rules> {

<Basic information>::=(number, address)

<Material Reserve Information>::=(Material Type, Quantity of Materials)

<Integrated traffic network information>::=(railway accessibility, road accessibility, air accessibility, single maximum transport capacity)

<bid rules>::=(historical utility library)

}

Based on the above information, the rescue point agent creates a bid file as shown in Fig. 6. The bid documents containing subcontracting/joint activities should be detailed in the Note column.

Fig. 6

Rescue point agent bidding file format.

3.3.3 The main behavior of rescue point agent

(1) Apply for bidding

The rescue point agent continuously collects various bidding documents issued by the demand point agent in the system. First, according to the formula (5) of 3.3.3, the benefit (ie, the utility function value) generated by participating in the bid is estimated. If the benefit value is greater than the generalized time cost incurred in participating in the task (ie, satisfying formula (10)), then the bid is involved. $E_{D_{i}}^{M_{j}} - G_{S_{i}}^{M_{j}} \geq 0$ (10)

(2) Generate integrated traffic scheduling plan

After the bid is determined, the rescue point agent is based on the comprehensive dynamic road network and the accessibility of each mode of transportation, and plans the transportation plan from the emergency material reserve point to the emergency material distribution center according to formula (7). The utility function estimate is determined by 3.3.3 (10), and the calculation formula of cost in the generalized time cost is as shown in equations (12) and (13), respectively. $max (E_{D_{i}^{'}}^{M_{j}^{'}} - G_{S_{i}}^{M_{j}^{'}})$ (11) $C = \sum_{j} \sum_{k} q_{j}^{m} c_{ij}^{k} d_{ij}^{k} x_{ijk} + \sum_{k} \sum_{l} q_{j}^{m} z c_{i}^{kl} y_{ikl}$ (12) $T = \sum_{j} \sum_{k} q_{j}^{m} \frac{d_{ij}^{k}}{v_{ij}^{k}} x_{ijk}$ (13)

The formulas (12) and (13) means that the rescue point agent intelligently supplies the quantity of resources in the m-th position to the demand point j; $c_{ij}^{k}$ is the transportation cost of the unit distance of the kth transportation vehicle from the node i to the node j; ${zc}_{i}^{kl}$ is the unit transfer fee for the emergency rescue material to be transferred from the kth transportation mode to the first transportation mode at the node i; $t_{i}^{kl}$ is the time at which the goods are transferred from the kth transport to the first transport at node i; $d_{ij}^{k}$ is the transportation distance i of the kth transportation mode from the node i to the node j for the emergency materials; $v_{ij}^{k}$ is the transport speed of the kth cargo selected from the node i to the node j.

(3) Task subcontracting

After the rescue point agent wins the bid, the task is further decomposed under the premise of ensuring that the agent completes the maximum proportion of the task, and the remaining task amount is subcontracted to other rescue point agents. At this point, the rescue point agent will perform the workflow of the demand point agent as the “tenderer” and bid for other rescue points in the system. The winning point of the winning bid agent is only allowed to be subcontracted once.

(4) Forming a joint venture

For the bidding task issued by the demand point agent, multiple rescue points can form a “joint bidding body”, negotiate the amount of tasks undertaken by each agent, and bid. The optimal task allocation scheme between the rescue points in the joint venture should satisfy the formula (14). n is the number of rescue points in the joint venture. $max \sum_{n} (E_{D_{i}^{'}}^{M_{j}^{'}} - G_{S_{i}}^{M_{j}^{'}})$ (14)

3.4 Algorithm

Through the above improved multi-agent game rules, a multi-agent-based trans-region integrated traffic emergency dispatch task assignment method can be established. The algorithm steps are as follows:

Step 1: Locate the location of the demand point and obtain information such as the accident level, accident casualty, and the start and end point of the accident;

Step 2: Determine the number of resource types required by the demand point and the storage amount of the rescue point;

Step 3: Set the rescue point set, S ={ S₁, S₂, ⋯ , S_N }, set the demand points, D ={ D₁, D₂, ⋯ , D_N };

Step 4: According to the bidding result, select m suitable rescue points, and let i = 1;

Step 5: According to the congestion degree and accessibility of various transportation modes from the rescue point to the demand point, generate n transportation combinations for the selected rescue points, and let j = 1;

Step 6: Obtain the optimal path of the ith resource point in the jth transport mode. If the ith resource supply point is reachable, let j = j+1, otherwise return to step 1;

Step 7: If j > n, let i = i+1, otherwise return to step 3;

Step 8: If i< =m, return to step 2. If i > m, the resource scheduling scheme under each transport combination mode is obtained, and the algorithm ends.

4 Experimental simulation

4.1 Simulation model parameter type

The model parameter types are as shown in Table 1.

Table 1
Input parameter types of the simulation model

Agent Parameter

Rescue point Geographic location

Resource category

Quantity of goods

Optional mode of transport

Transportation parameters (unit cost, time, etc.)

Rescue point Geographic location

Whether trans-region

Resource type

Number of resources

Optional mode of transport

Transportation parameters (unit cost, time, etc.)

Agent	Parameter
Rescue point	Geographic location
	Resource category
	Quantity of goods
	Optional mode of transport
	Transportation parameters (unit cost, time, etc.)
Rescue point	Geographic location
	Whether trans-region
	Resource type
	Number of resources
	Optional mode of transport
	Transportation parameters (unit cost, time, etc.)

4.2 Simulation case

To verify the validity of the model, we take the case of an oil storage tank explosion in Qingdao, Shandong Province, China. After the incident, the demand point agent releases real-time demand information, which requires 495 units of Class A resources, 845,000 units of Class B resources, and 75 units of Class C resources.

The parameters such as the location of the rescue point agent and the type and amount of resources stored are as Table 2:

Table 2
Rescue point agent information table

Resource category Class A Class A Class B Class B Class C

Rescue unit Shandong Reserve Resources Administration 832 Emergency Rescue Resource Warehouse Emergency Supplies Supply point Integrated Emergency Rescue Team Henan Reserve

Rescue point address 80 Jingcheng Road, Qingdao, Shandong, China 200 meters east of Taoyuan South Street, Dongcheng District, Beijing 835 Guolan Street, Chengdu, Sichuan Province No. 28, Health Road, Kashi City, Xinjiang Uygur Autonomous Region No.179 Dongming Road, Jinshui District, Zhengzhou City

Bidding form Demand point invitation tender Rescue point independent bidding Rescue point independent bidding Rescue point independent bidding Rescue point independent bidding

City Qingdao Beijing Beijing Chengdu Zhengzhou

Mode of transport Highway Railway Railway, Xi’an transit, highway Railway, Urumqi transit, aviation Highway, Jinan transit, railway

Distance/km 27.3 710 609.8, 1173.9 1316, 2904.4 454.4, 318

Time/h 0.68 9.467 23.4, 47 26.3, 97 11.5, 8

Actual supply/unit 210 285 225000 620000 75

Total cost/10,000 yuan 6573 103141.5 22953.15 222769.1 57385.5

Resource category	Class A	Class A	Class B	Class B	Class C
Rescue unit	Shandong Reserve Resources Administration 832	Emergency Rescue Resource Warehouse	Emergency Supplies Supply point	Integrated Emergency Rescue Team	Henan Reserve
Rescue point address	80 Jingcheng Road, Qingdao, Shandong, China	200 meters east of Taoyuan South Street, Dongcheng District, Beijing	835 Guolan Street, Chengdu, Sichuan Province	No. 28, Health Road, Kashi City, Xinjiang Uygur Autonomous Region	No.179 Dongming Road, Jinshui District, Zhengzhou City
Bidding form	Demand point invitation tender	Rescue point independent bidding	Rescue point independent bidding	Rescue point independent bidding	Rescue point independent bidding
City	Qingdao	Beijing	Beijing	Chengdu	Zhengzhou
Mode of transport	Highway	Railway	Railway, Xi’an transit, highway	Railway, Urumqi transit, aviation	Highway, Jinan transit, railway
Distance/km	27.3	710	609.8, 1173.9	1316, 2904.4	454.4, 318
Time/h	0.68	9.467	23.4, 47	26.3, 97	11.5, 8
Actual supply/unit	210	285	225000	620000	75
Total cost/10,000 yuan	6573	103141.5	22953.15	222769.1	57385.5

According to the algorithm given in Section 2, do the following steps:

The demand point agent releases the bidding information according to the type of demand materials and the sub-segment;

The demand point agent invites the target to the rescue point of the same region (ie, within Shandong Province). Out-of-the-box agents in the region need to maximize the transportation of demanded goods;

Under the premise that the demand point agent in the same region meets the material demand to the maximum extent, calculate the benefit value and the generalized time cost of each rescue point agent outside the region participating in the task;

Screening all the schemes that conform to formula (10), and sorting them in descending order according to the size of the values;

Add the transportation amount given by each scheme in turn until the demand is met. The rescue point agent distributed throughout the country rescues according to the emergency dispatching rules, and finally concludes that the comprehensive traffic emergency dispatching plan is shown in Table 3. The schematic diagram of the scheme is shown in Fig. 8.

Table 3

Simulation experiment results

Serial number	Rescue unit	Rescue point address	City	Resource type	Storage
1	Shandong Reserve Materials Administration 832	80 Jingcheng Road, Qingdao, Shandong, China	Qingdao	Class A	210
2	State Material Reserve Bureau (Tianjin 833)	Tianjin Binhai New region Haibin Expressway	Tianjin	Class A	157.5
3	Emergency rescue material warehouse	200 meters east of Taoyuan South Street, Dongcheng District, Beijing	Beijing	Class A	300
4	133 State Material Reserve Bureau	G307, Luquan District, Shijiazhuang City, Hebei Province	Shijiazhuang	Class B	150000
5	Henan Reserve Materials Administration	No.179 Dongming Road, Jinshui District, Zhengzhou City	Zhengzhou	Class B	150000
6	Emergency supplies supply point	835 Guolan Street, Chengdu, Sichuan Province	Chengdu	Class B	300000
7	Integrated emergency rescue team	No. 28, Health Road, Kashi City, Xinjiang Uygur Autonomous Region	Xinjiang	Class B	900000
8	Shandong Reserve Materials Administration 832	80 Jingcheng Road, Qingdao, Shandong, China	Qingdao	Class C	30
9	Emergency supplies warehouse	Near No. 39, Jiqi Road, Tianqiao District, Jinan City, Shandong Province	Jinan	Class C	75
10	133 State Material Reserve Bureau	G307, Luquan District, Shijiazhuang City, Hebei Province	Shijiazhuang	Class C	75
11	Henan Reserve Materials Administration	No.179 Dongming Road, Jinshui District, Zhengzhou City	Zhengzhou	Class C	120

Fig. 7

Task subcontracting and consortium venture.

Fig. 8

Schematic diagram of the simulation scheme.

5 Conclusions

In this paper, based on the characteristics of trans-regional integrated traffic emergency dispatch, an improved multi-agent model based on game rules was introduced to distribute emergency dispatching tasks. The experimental simulation showed that the model can simulate the dynamic decision of the rescue point and demand point in the emergency dispatching process, which had certain reference.

The main contributions of this paper are as follows:

Introducing a multi-agent system, allowing the rescue point and demand point to communicate and coordinate the emergency material dispatching plan independently, thus replacing the role of the traditional “Central Controller” and reducing the system operation and operation dimension;

Through the bidding rules, mobilizing the initiative and enthusiasm of each rescue point for the rescue of the rescue body, so as to realize the trans-regional dispatch of emergency materials;

Based on the comprehensive transportation network, considering multiple modes of transportation to achieve multimodal transport of emergency materials.

Footnotes

Acknowledgments

This research was supported by the National Key R&D Program of China (2016YFC0802208), National Natural Science Foundation of China (Project No. 61703351, 71971182) , Sichuan Science and Technology Program (Project NO. 2018RZ0078, 2019JDR0211), Science and Technology Plan of China Railway Corporation (Project No.: P2018T001, K2018X012, N2018X006-01).

References

Wenhua

Zhao

and Yue

Yan

, Emergency rescue, Beijing National Defense University Press (2015), 213–217.

Wangyang

, A guide to emergency response plans for public emergencies, Beijing Higher Education Press (2007), 9.

Nadjla GhaemiEmail CatsRob authorOded and Goverde

M.P.

, Railway disruption management challenges and possible solution directions, Public TransportJuly9(1–2) (2017), 343–364.

Cheng

, Research on the location and scheduling methods of multi-resource and multi-supplier emergency problem, Xian University of Electronic Technology, 2014.

Xueying

Gao

, Research on the optimization methods of large-scale emergency rescue resources layout and scheduling, Jilin University, 2012.

Bodaghi

Behrooz

and Palaneeswaran

Ekambaram

, An optimization model for scheduling emergency operations with multiple teams, Proceedings of the 2016 International Conference on Industrial Engineering and Operations Management.

Qing

Chang

, Vehicle schedulingmodel of emergency logistics distribution based on internet of things, International Journal of Applied Decision Sciences, 2018.

Rawls

C.G.

and Turnquist

M.A.

, Pre-positioning and dynamic delivery planning for short-term response following a natural disaster, Socio-Economic Planning Science46(1) (2012), 46–54.

Barbarosoglu

, Özdamar

and Çevik

, An interactive approach for hierarchical analysis of helicoter logistics in disater relief operations, European Journal of Opeational Research140(1) (2002), 118–133.

10.

Chang

M.S

, Tseng

Y.L.

and Chen

J.W.

, A scenario planning approach for the flood emergency logistics praparation problem under uncertainty, Transportation Research Part E Logistics & Transportation Review43(6) (2012), 737–754.

11.

Torreno

Alejandro

, Onaindia

Eva

and Sapena

Oscar

, Fmap: Distributed cooperative multi-agent planning, Applied Intelligence41(2) (2014), 606–626.

12.

ter Mors

Adriaan

, Yadati

Chetan

, Witteveen

Cees

and Zhang

Yingqian

, Coordination by design and the price of autonomy, Autonomous Agents and Multi-Agent Systems20(3) (2010), 308–341.

13.

Yuming

Lin

, Highway emergency management and assistant decision making, Beijing: Intellectual Property Publishing House (2014), 34.

14.

Torabi

Behnam

, Wenkstern

Rym Z.

and Saylor

Robert

, A self-adaptive collaborative multi-agent based raffiTc signal timing system, 2018 IEEE International Smart Cities Conference (ISC2).

15.

Yan

, Shengwei

XING

and Yingjun

ZHANG

, Maritime search task allocation with multi-agent concept, Navigation of China41(03) (2018), 91–94+100.

16.

, Guolei

Tang

, Xiangqun1

Song

, Jingjing

, Yue2

, Xuhui

, Xiaoyi1

Zhao

and Yonghao

Zhang

, Multi-agent simulation modeling for railway loading/unloading line of container terminal, Port Engineering Technology55(04) (2018), 6–9.

17.

Dingjun

Chen

, Xu

Chang’an

and Ni

Shaoquan

, Data mining on Chinese train accidents to derive associated rules, Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit231(2) (2017), 239–252.

18.

Chen

Chien-Ming

, Wang

King-Hang

, Yeh

Kuo-Hui

, Xiang

Bin

and Wu

Tsu-Yang

, Attacks and solutions on a three-party password-based authenticated key exchange protocol for wireless communications, Journal of Ambient Intelligence and Humanized Computing10(8) (2019), 3133–3142.

19.

Ming-Tai Wu

Jimmy

, Chun-Wei Lin

Jerry

and Tamrakar

Ashish

, High-utility itemset mining with effective pruning strategies,ACMTransactions on Knowledge Discovery from Data, 2019 https://doi.org/10.1145/3363571

20.

Ming

Cao

and Weiguo

Fang

, Distributed MMAS for weapon target assignment based on spark framework, Journal of Intelligent & Fuzzy Systems35(3) (2018), 3685–3696.

21.

Rahaie

Zahra

and Beigy

Hamid

, Critic learning in multi agent credit assignment problem, Journal of Intelligent & Fuzzy Systems30 (2016), 3465–3480.

22.

Zhongzhi

, Zhang

Qingpeng

, Dingjun

Chen

and Yuxin

, Characterizing the connectivity of railway networks [J], IEEE Transactions on Intelligent Transportation Systems (2019), 1–12. DOI: 10.1109/TITS.2019.2909120.