Sensor distribution and scheduling method for multi-vehicle-type hybrid mobile sensing

Abstract

The invention discloses a sensor distribution and scheduling method for multi-vehicle-type hybrid mobile perception, and the method specifically comprises the steps: proposing a quantitative index of perception capability, and quantifying the perception capability of a motorcade through a space-time coverage effect; calculating the expected coverage of each vehicle type on the grid, and representing the average coverage frequency of the taxi by using the expectation of binomial distribution; for the buses, calculating the average coverage frequency of the bus fleet configured with the sensors to each grid; a route planning algorithm is adopted for the special vehicle; and establishing a sensor distribution model for three vehicle types, and solving by adopting an iterative solution algorithm. Quantitative analysis can be carried out according to different sensing requirements, and target sensing distribution is deduced through first-order analysis; the problem of how to allocate limited cost to three types of vehicles (taxis, buses and special vehicles) is solved, and bus lines are selected and special monitoring vehicles are scheduled, so that optimal space-time coverage is provided for a monitoring area.

Description

Technical Field

[0001] The present invention belongs to mobile sensing technology, and in particular to a sensor allocation and scheduling method for mixed mobile sensing of multiple vehicle types. Background Art

[0002] With the advent of the smart city era, the popularity of various sensor devices (such as smartphones and low-cost wireless sensors) has given rise to a data collection model based on crowdsourcing perception. Among the many crowdsourcing perception methods, vehicle-based mobile perception (also known as vehicle-mounted perception) has gradually attracted widespread attention due to its low deployment cost, strong mobility and wide spatial and temporal coverage. This technology has been applied in many fields of smart cities, including air quality monitoring, traffic status estimation, noise pollution monitoring and urban infrastructure health monitoring.

[0003] In cities, the three common types of perception vehicles (taxis, buses, and special vehicles) have their own advantages and disadvantages in terms of spatial coverage, temporal coverage, controllability, reliability, and economy due to their different mobility modes. Existing research mostly focuses on the perception application of a single vehicle type (such as taxis or buses), but this homogeneous fleet approach has certain limitations. Summary of the invention

[0004] The present invention aims to solve the problem of sensor allocation in multi-vehicle hybrid mobile perception, so as to better make up for the shortcomings of single vehicle monitoring and improve perception efficiency. To this end, the present invention provides a sensor allocation and scheduling method for multi-vehicle hybrid mobile perception.

[0005] A sensor allocation and scheduling method for mixed mobility perception of multiple vehicle types according to the present invention comprises the following steps:

[0006] Step 1: Propose quantitative indicators of perception ability.

[0007] The perception capability of the fleet is quantified by the temporal and spatial coverage effect. Specifically, the monitoring area is first discretized into a group of grids Discretize the monitoring time range into a set of time windows A group of spatiotemporal perception units are formed, which are quantified by counting the number of vehicles covering each spatiotemporal grid. As the number of covered vehicles increases, the perception benefit of this grid increases. At the same time, in order to avoid vehicles from crowding in certain grids and scoring frequently, the quantization function also needs to satisfy the diminishing marginal gain, as shown below:

[0008]

[0009] In the formula, g represents the index of the spatial unit; t represents the index of the time unit; N g,trepresents the sample size collected by the sensor fleet in the spatiotemporal unit (g, t); β is the utility function parameter. The smaller β is, the smaller the marginal utility of multiple sampling is; ξ(·) represents the sensor utility function.

[0010] Finally, according to different monitoring needs, time weights and space weights are introduced and normalized, and the weighted average scores of all grids within the entire research time range are used to quantify the perception effect, which is understood as the "average benefit in time and space", as shown below:

[0011]

[0012] Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t, and satisfies w g Represents the perceptual weight corresponding to the spatial unit g, and satisfies

[0013] This index is used as the optimization target for the subsequent sensor allocation problem.

[0014] Step 2: Calculate the expected coverage of the grid for each vehicle type.

[0015] The optimization goal of the mobile sensing coverage problem of multiple vehicle models is to maximize the average spatiotemporal benefit, which is calculated by the number of vehicles covering each spatiotemporal grid. The number of vehicles covering each grid is then the cumulative value of the coverage times of three different vehicle models. The coverage times of each vehicle model is affected by the cost assigned to it. Therefore, we first solve how to calculate the coverage times of each spatiotemporal grid for each vehicle model at a given cost.

[0016] For taxis, we use the actual taxi trajectory data to verify that the coverage of the grid by taxis follows a binomial distribution. Therefore, given the number of taxis n T and the binomial distribution parameter p for each space-time grid obtained by fitting the actual data g,t After that, the average number of times a taxi covers the grid is expressed as the expectation of the binomial distribution, as follows:

[0017]

[0018] In the formula, represents the expected number of times sensor-equipped taxis cover the spatiotemporal grid (g, t); n T represents the number of taxis equipped with sensors; p g,t is the binomial distribution parameter of the space-time grid (g, t).

[0019] For buses, an analytical model is used to calculate the average number of times a bus fleet equipped with sensors covers each grid. Specifically, it is calculated by the number of sensors equipped on the route × the average number of coverage times per hour × the service intensity parameter. The specific calculation expression is as follows:

[0020]

[0021] In the formula, represents the expected number of times that a bus equipped with sensors covers the spatiotemporal grid (g, t); m represents the number of alternative bus routes; δ gj is a 0-1 parameter, indicating whether the j-th bus line can cover the spatial grid g; γ j (t) is the service intensity parameter of the jth bus line in the tth time period; represents the number of sensors installed on the j-th bus route; represents the single service time of bus route j; represents the turnaround time of bus line j at the terminal.

[0022] For special vehicles, a greedy heuristic algorithm or a meta-heuristic algorithm is designed according to the target distribution to plan the route R of each special vehicle. D , and then count the number of times each space-time grid is covered based on the route planning results

[0023] Step 3: Solve the sensor allocation problem for three types of vehicles.

[0024] Combined with the calculation method of the number of times taxis, buses, and special vehicles cover the spatiotemporal grid in step 2, the sensor allocation problem for the three types of vehicles is described as the following model:

[0025]

[0026] Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t; w g Represents the perceptual weight corresponding to the spatial unit g; N g,t represents the expected coverage times of the spatiotemporal grid (g, t) by the sensor fleet; β is the utility function parameter; They represent the expected coverage times of the taxi sensor fleet, bus sensor fleet, and dedicated sensor fleet on the spatiotemporal grid (g, t) respectively; c T ,c B ,c D Respectively represent the unit cost of deploying sensors on a taxi, bus, and special vehicle; M represents the total budget; n Tis an integer decision variable, indicating the number of sensors deployed in the taxi fleet; is an integer decision variable, indicating the number of sensors deployed on the jth bus line; n D is an integer decision variable, representing the number of special vehicles.

[0027] The model is solved by an iterative solution algorithm. The hybrid mobile sensor coverage problem of taxis and buses is regarded as a convex optimization problem and solved quickly using a solver. Furthermore, the route planning algorithm for special vehicles is incorporated into the solution framework of hybrid monitoring. This problem is regarded as a two-level planning problem. The upper problem is to allocate the given cost to the three types of vehicles, and the lower problem is the convex optimization model of taxis and buses and the route planning problem of special vehicles. After determining the cost of installing sensors for each type of vehicle based on actual operations, all possible special vehicle deployment plans are traversed in turn. When the budget for the special vehicle is determined, the two sub-problems of the lower level are solved iteratively to determine the optimal sensor deployment method.

[0028] The beneficial technical effects of the present invention are:

[0029] (1) The present invention proposes a new perception effect evaluation function, which can perform quantitative analysis according to different perception requirements and derive the target perception distribution through first-order analysis.

[0030] (2) The present invention proposes a two-layer optimization framework to solve the problem of how to allocate limited costs to three types of vehicles (taxi, bus, and special-purpose vehicle), and to select bus lines and dispatch special-purpose monitoring vehicles, so as to provide the best spatiotemporal coverage for the monitoring area. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram of the process flow of the sensor allocation and scheduling method for hybrid mobility perception of multiple vehicle types according to the present invention. DETAILED DESCRIPTION

[0032] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0033] The process of the sensor allocation and scheduling method for multi-vehicle hybrid mobility perception of the present invention is as follows: Figure 1 As shown, the specific steps include:

[0034] Step 1: Propose quantitative indicators of perception ability.

[0035] The perception capability of the fleet is quantified by the temporal and spatial coverage effect. Specifically, the monitoring area is first discretized into a group of grids Discretize the monitoring time range into a set of time windows A group of spatiotemporal perception units is formed, and the number of vehicles covering each spatiotemporal grid is counted to quantify; as the number of covered vehicles increases, the perception benefit of this grid increases. At the same time, in order to avoid vehicles from crowding in certain grids and scoring frequently, the quantization function also needs to satisfy the diminishing marginal gain (in the present invention, the power function is used for quantification), as shown below:

[0036]

[0037] In the formula, g represents the index of the spatial unit; t represents the index of the time unit; N g,t represents the sample size collected by the sensor fleet in the spatiotemporal unit (g, t); β is the utility function parameter. The smaller β is, the smaller the marginal utility of multiple sampling is; ξ(·) represents the sensor utility function.

[0038] Finally, according to different monitoring needs, time weights and space weights are introduced and normalized, and the weighted average scores of all grids within the entire research time range are used to quantify the perception effect, which is understood as the "average benefit in time and space", as shown below:

[0039]

[0040] Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t, and satisfies w g Represents the perceptual weight corresponding to the spatial unit g, and satisfies

[0041] This indicator is used as the optimization target for the subsequent sensor allocation problem. By analyzing the KKT conditions of the optimization model, it is found that this optimization target, while maximizing the average spatiotemporal benefit, also indirectly optimizes the difference between the actual distribution of vehicle coverage times and the target distribution.

[0042] Step 2: Calculate the expected coverage of the grid for each vehicle type.

[0043] The optimization goal of the mobile sensor coverage problem of multiple models is to maximize the average spatiotemporal benefit, which is calculated by the number of vehicles covering each spatiotemporal grid. The number of vehicles covering each grid is then the cumulative value of the coverage times of three different models. The coverage times of each model is affected by the cost assigned to it. Therefore, we first solve how to calculate the coverage times of each spatiotemporal grid for each model at a given cost.

[0044] For taxis, we use the actual taxi trajectory data to verify that the coverage of the grid by taxis follows a binomial distribution. Therefore, given the number of taxis n T and the binomial distribution parameter p for each space-time grid obtained by fitting the actual datag,t After that, the average number of times a taxi covers the grid is expressed as the expectation of the binomial distribution, as follows:

[0045]

[0046] In the formula, represents the expected number of times sensor-equipped taxis cover the spatiotemporal grid (g, t); n T represents the number of taxis equipped with sensors; p g,t is the binomial distribution parameter of the space-time grid (g, t).

[0047] For buses, an analytical model is used to calculate the average number of times a bus fleet equipped with sensors covers each grid; specifically, it is calculated by the number of sensors equipped on the line × the average number of coverage times per hour × the service intensity parameter. The service intensity parameter is introduced mainly because the number of vehicles running on the line will fluctuate over time. For example, during peak hours, all vehicles on the line will be used, but during off-peak hours, some vehicles will be randomly idle. If the vehicles equipped with sensors are idle, the coverage quality will decrease. Therefore, the service intensity parameter is used to quantify the impact of this factor. The specific calculation expression is as follows:

[0048]

[0049] In the formula, represents the expected number of times that a bus equipped with sensors covers the spatiotemporal grid (g, t); m represents the number of alternative bus routes; δ gj is a 0-1 parameter, indicating whether the j-th bus line can cover the spatial grid g; γ j (t) is the service intensity parameter of the jth bus line in the tth time period; represents the number of sensors installed on the j-th bus route; represents the single service time of bus route j; represents the turnaround time of bus line j at the terminal.

[0050] For special vehicles, their operation is completely controllable, so a greedy heuristic algorithm or a meta-heuristic algorithm is designed according to the target distribution to plan the route R of each special vehicle. D , and then count the number of times each space-time grid is covered based on the route planning results

[0051] Step 3: Solve the sensor allocation problem for three types of vehicles.

[0052] Combined with the calculation method of the number of times taxis, buses, and special vehicles cover the spatiotemporal grid in step 2, the sensor allocation problem for the three types of vehicles is described as the following model:

[0053]

[0054]

[0055] Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t; w g Represents the perceptual weight corresponding to the spatial unit g; N g,t represents the expected coverage times of the spatiotemporal grid (g, t) by the sensor fleet; β is the utility function parameter; They represent the expected coverage times of the taxi sensor fleet, bus sensor fleet, and dedicated sensor fleet on the spatiotemporal grid (g, t) respectively; c T ,c B ,c D Respectively represent the unit cost of deploying sensors on a taxi, bus, and special vehicle; M represents the total budget; n T is an integer decision variable, indicating the number of sensors deployed in the taxi fleet; is an integer decision variable, indicating the number of sensors deployed on the jth bus line; n D is an integer decision variable, representing the number of special vehicles.

[0056] The model is solved by an iterative solution algorithm. The hybrid mobile sensor coverage problem of taxis and buses is regarded as a convex optimization problem and solved quickly using a solver. Furthermore, the route planning algorithm for special vehicles is incorporated into the solution framework of hybrid monitoring. This problem is regarded as a two-level planning problem. The upper problem is to allocate the given cost to the three types of vehicles, and the lower problem is the convex optimization model of taxis and buses and the route planning problem of special vehicles. After determining the cost of installing sensors for each type of vehicle based on actual operations, all possible special vehicle deployment plans are traversed in turn. When the budget for the special vehicle is determined, the two sub-problems of the lower level are solved iteratively to determine the optimal sensor deployment method.

[0057] In Longquanyi District, Chengdu (557km 2 ) implementation case, hybrid mobile sensing based on multiple vehicle types (taxis, buses, special vehicles) can save 39%-61% of the total cost compared to single vehicle type sensing; and under the same cost, the utility of hybrid mobile sensing can be improved by 11%-21%.

Claims

1. A sensor allocation and scheduling method for multi-vehicle hybrid mobility perception, characterized in that: The following steps are involved: Step 1: Propose quantitative indicators of perception ability; The perception capability of the fleet is quantified by the temporal and spatial coverage effect. Specifically, the monitoring area is first discretized into a group of grids Discretize the monitoring time range into a set of time windows A group of spatiotemporal perception units are formed, which are quantified by counting the number of vehicles covering each spatiotemporal grid. As the number of covered vehicles increases, the perception benefit of this grid increases. At the same time, in order to avoid vehicles from crowding in certain grids and scoring frequently, the quantization function also needs to satisfy the diminishing marginal gain, as shown below: In the formula, g represents the index of the spatial unit; t represents the index of the time unit; N g,t represents the sample size collected by the sensor fleet in the spatiotemporal unit (g, t); β is the utility function parameter. The smaller β is, the smaller the marginal utility of multiple sampling is; ξ(·) represents the sensor utility function; Finally, according to different monitoring needs, time weight and space weight are introduced and normalized, and the weighted average score of all grids within the entire research time range is used to quantify the perception effect, which is understood as the "average benefit in time and space", as shown below: Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t, and satisfies w g Represents the perceptual weight corresponding to the spatial unit g, and satisfies This index is used as the optimization target for the subsequent sensor allocation problem; Step 2: Calculate the expected coverage of the grid for each vehicle type; The optimization goal of the mobile sensing coverage problem of multiple models is to maximize the average spatiotemporal benefit, which is calculated by the number of vehicles covering each spatiotemporal grid. The number of vehicles covering each grid is the cumulative value of the number of times three different models cover each grid. The number of times each model covers each grid is affected by the cost assigned to it. Therefore, we first solve how to calculate the number of times each model covers each spatiotemporal grid at a given cost. For taxis, we use the actual taxi trajectory data to verify that the coverage of the grid by taxis follows a binomial distribution. Therefore, given the number of taxis n T and the binomial distribution parameter p for each space-time grid obtained by fitting the actual data g,t After that, the average number of times a taxi covers the grid is expressed as the expectation of the binomial distribution, as follows: In the formula, represents the expected number of times sensor-equipped taxis cover the spatiotemporal grid (g, t); n T represents the number of taxis equipped with sensors; p g,t is the binomial distribution parameter of the space-time grid (g, t); For buses, an analytical model is used to calculate the average number of times a bus fleet equipped with sensors covers each grid. Specifically, it is calculated by the number of sensors equipped on the route × the average number of coverage times per hour × the service intensity parameter. The specific calculation expression is as follows: In the formula, represents the expected number of times a bus equipped with sensors covers the spatiotemporal grid (g, t); m represents the number of alternative bus routes; δ gj is a 0-1 parameter, indicating whether the j-th bus line can cover the spatial grid g; γ j (t) is the service intensity parameter of the jth bus line in the tth time period; represents the number of sensors installed on the j-th bus line; represents the single service time of bus route j; represents the turnaround time of bus route j at the terminal; For special vehicles, a greedy heuristic algorithm or a meta-heuristic algorithm is designed according to the target distribution to plan the route R of each special vehicle. D , and then count the number of times each space-time grid is covered based on the route planning results Step 3: Solve the sensor allocation problem for three types of vehicles; Combined with the calculation method of the number of times taxis, buses, and special vehicles cover the spatiotemporal grid in step 2, the sensor allocation problem for the three types of vehicles is described as the following model: Where Φ represents the total perceived utility score of the perceived fleet; μ t represents the perceptual weight corresponding to the time unit t; w g Represents the perceptual weight corresponding to the spatial unit g; N g,t represents the expected coverage times of the spatiotemporal grid (g, t) by the sensor fleet; β is the utility function parameter; They represent the expected coverage times of the taxi sensor fleet, bus sensor fleet, and dedicated sensor fleet on the spatiotemporal grid (g, t) respectively; c T ,c B ,c D Respectively represent the unit cost of deploying sensors on a taxi, bus, and special vehicle; M represents the total budget; n T is an integer decision variable, indicating the number of sensors deployed in the taxi fleet; is an integer decision variable, indicating the number of sensors deployed on the jth bus line; n D is an integer decision variable, indicating the number of special vehicles; The model is solved by an iterative solution algorithm. The hybrid mobile sensor coverage problem of taxis and buses is regarded as a convex optimization problem and solved quickly using a solver. Furthermore, the route planning algorithm for special vehicles is incorporated into the solution framework of hybrid monitoring. This problem is regarded as a two-level planning problem. The upper problem is to allocate the given cost to the three types of vehicles, and the lower problem is the convex optimization model of taxis and buses and the route planning problem of special vehicles. After determining the cost of installing sensors for each type of vehicle based on actual operations, all possible special vehicle deployment plans are traversed in turn. When the budget for the special vehicle is determined, the two sub-problems of the lower level are solved iteratively to determine the optimal sensor deployment method.

Images