CN105788334A - Urban path finding method taking personal preferences of drivers into consideration - Google Patents

Abstract

The invention provides an urban path finding method taking personal preferences of drivers into consideration. The method comprises the following steps: collecting driving GPS data of a driver and map attribute data to get a specific path chosen by the driver in previous travels and the information of the path including the road grade, length and the number of passing intersections; defining a road section generalized cost containing a parameter Beta and a multinomial Logit model to get the probability that the driver chooses the path, and estimating the personal preference parameter Beta through maximum likelihood estimation; and finding a path in conformity with the driver's preference and with the lowest generalized cost by use of a Dijkstra shortest path algorithm according to the calibrated generalized road section cost. The method has the advantage that the usual theoretical experience value is not taken for the measured impedance of a road section, and a shortest path is offered to a driver on the premise of considering the personal preference of the driver according to historical driving paths and on the basis of redefining the road section generalized cost.

Description

A kind of city path finding method considering driver individual's preference

Technical field

The present invention relates to a kind of method finding shortest path, be specifically related to a kind of personalized city path finding method considering driver individual's preference.

Background technology

Developing rapidly and extensive use of information science technology, drive whole society's demand to spatial information, it is a kind of under computer hardware and software is supported that GIS-Geographic Information System (is called for short GIS), based on spatial database, use the theory of system engineering and information science, spatial data is carried out scientific management and comprehensive analysis, for the technological system of planning, decision, management and research offer information.

Analysis of network, as one of topmost function of GIS, has played important effect in electronic navigation, traffic for tourism, urban planning, and in analysis of network, the problem of most basic most critical has been shortest route problem.It, as the basis selecting optimal problem in many fields, occupies critical role in traffic network analysis system.From the angle of network model, Shortest Path Analysis is exactly look for a path hindering intensity minimum in specifying network between two nodes.Shortest route problem is always up a study hotspot of the subject such as traffic engineering, Geomatics, and computer data structure and effectively combining of algorithm that classical graph theory is perfect with development make new shortest path first continue to bring out, and differ from one another.Shortest Path Analysis is usually used in auto-navigation system and various City Integrated Emergency Response System etc. in practice, for instance at driving conditions, will calculate vehicle front travel route etc. in real time.

Summary of the invention

Technical problem: the present invention provides the city path finding method considering driver individual's preference that can meet personal like, personalized path finding.

Technical scheme: the city path finding method considering driver individual's preference of the present invention, comprises the following steps:

1) obtain driver trip historical data, described historical data be in the trip that driver is former select concrete path and the category of roads in path, length, by crossing quantity；

2) calculate the generalized cost V in section as follows, and it can be used as actual measurement impedance:

$c_k^{rs}=V={\mathrm{\β}}_1{l}_{1}a+{\mathrm{\β}}_2{l}_{2}b+{\mathrm{\β}}_3{l}_{3}c+{\mathrm{\β}}_4*1$

Wherein, k is certain paths selected, and r is starting point, and s is settled point,For the actual measurement impedance of path k between starting point r and settled some s；V is section generalized cost；Whether a is be highway: is that highway takes 1, is not that highway takes 0；Whether b is be major trunk roads: is that major trunk roads take 1, is not that major trunk roads take 0；Whether c is be branch road: is that branch road takes 1, is not that branch road takes 0；l₁For express highway section length, l₂For turnpike road segment length, l₃For branch road road section length；β₁For express highway section weight coefficient, β₂For major trunk roads section weight coefficient, β₃For branch road section weight coefficient, β₄For crossing weight coefficient；Wherein β₄* 1 for considering the number in section, the impact that namely expense is produced by the number of crossing；

Described β₁、β₂、β₃、β₄All try to achieve as follows: based on multinomial Logit mode, try to achieve the probability of selecting paths k；Based on historical data, estimate express highway section weight coefficient, major trunk roads section weight coefficient, branch road section weight coefficient, crossing weight coefficient with maximum likelihood estimate respectively；

3) utilize Dijkstra shortest path first, find the path that the generalized cost meeting drivers preference is the shortest.

Further, in the inventive method, described step 2) in, obtain the probability of selecting paths k in such a way:

It is primarily based on logit model, solves according to following formula and select the probability of path k between starting point r and settled some s

$p_k^{rs}=\frac{\exp (-{\mathrm{\θc}}_k^{rs}/\stackrel{\‾}{c})}{\underset{l\∈R_{rs}}{\Σ}\exp (-{\mathrm{\θc}}_l^{rs}/\stackrel{\‾}{c})}$

Wherein,For the actual measurement impedance of path k between starting point r and settled some s,For the meansigma methods of all path impedance, θ is conversion parameter, R_rsBeing the set in all paths between starting point s and settled some r, l is certain paths in set of paths；

Then the probability of selecting paths k is calculated according to following formula:

$p_k^{rs}\left(\mathrm{\β}\right)=\frac{\exp \[-{\mathrm{\θc}}_k^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}{\underset{l\∈R_{rs}}{\Σ}\exp \[-{\mathrm{\θc}}_l^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}$

Wherein, β represents β₁、β₂、β₃Or β₄。

Further, in the inventive method, described step 2) in estimate express highway section weight coefficient, major trunk roads section weight coefficient, branch road section weight coefficient, crossing weight coefficient as follows:

Between the starting point r and the settled some s that are drawn by historical data, the probability P (β) of Path selection situation is:

P (β)=n₁logp₁(β)+n₂logp₂(β)+...+n_nlogp_n(β)

Wherein, β represents β₁、β₂、β₃Or β₄, n₁For the selected number of times in path 1, p₁(β) for the selected probability in path 1, n₂For the selected number of times in path 2, p₂(β) for the selected probability in path 2, n_nFor the selected number of times of path n, p_n(β) for the selected probability of path n；

Then β is estimated according to following log-likelihood equation group₁、β₂、β₃、β₄:

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_1}=0$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_2}=0$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_3}=0$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_4}=0$

Further, in the inventive method, described step 3) in Dijkstra shortest path first comprise the following steps:

Step0: initialize.

Step1: terminate inspection.

Step2: amendment T label.

Step3: determine P label.

Beneficial effect: the present invention compared with prior art, has the advantage that

Research commonly uses link travel time function and BPR function before; demarcate the generalized cost of road; perhaps this section is shortest path in theory; but the practical situation in conjunction with driver; there may be individual and have special demand, it will usually select a non-theoretic shortest path but a paths of individual's preference；The present invention then oneself definition section generalized cost (tried to achieve by the method mentioned in literary composition by the parameter beta in expense, and the method is exactly consider actual individual's preference of driver rather than existing theoretical hypothesis) so that the more hommization of the shortest path of offer, personalization.

The present invention is by the collection driving gps data and map attribute data to driver, demarcate broad sense Road Expense, in conjunction with multinomial Logit mode, maximal possibility estimation is used to estimate the generalized cost considering individual's parameter, utilize the shortest path that Dijkstra shortest path first obtains on this basis, it is then find the preference meeting driver, the path that generalized cost is the shortest.Advantage is in that the mensuration impedance in section is not take common theoretical empirical value, but according to history driving path, it is considered to driver individual's preference, having redefined, the basis of section generalized cost provides shortest path to driver.

Accompanying drawing explanation

Fig. 1 is the flow chart of the inventive method.

Fig. 2 is exemplary plot, has k between origin and destination₁, k₂......k_nPaths.

Detailed description of the invention

Below in conjunction with embodiment and Figure of description, the present invention is further illustrated.

The present invention provides a kind of personalized city path finding method considering driver individual's preference, is realized by following steps:

1) probability of selecting paths k is drawn by Logit path Choice Model.

In randomness Assignment Problem, between definition origin and destination r, s, each alternative path has gratifying degree is effectiveness, and traveler always selects the path that subjective judgment effectiveness is maximum, and the value of utility of kth paths can be expressed as:

$U_k=-c_k^{rs}=-{\mathrm{\θc}}_k^{rs}+{\mathrm{\ϵ}}_k^{rs}---\left(1\right)$

In formula, U_kEffectiveness for origin and destination selecting paths k；Actual measurement impedance for path；θ is by a positive transition parameter surveying impedance transformation effect；The stochastic error of the factors composition that expression path can not observe.

Logit model is based upon stochastic error, and on the basis that separate and obedience Gumbel is distributed, the probability solving selecting paths k based on logit model is:

$p_k^{rs}=\frac{\exp (-{\mathrm{\θc}}_k^{rs}/\stackrel{\‾}{c})}{\underset{l\∈R_{rs}}{\Σ}\exp (-{\mathrm{\θc}}_l^{rs}/\stackrel{\‾}{c})}---\left(2\right)$

In formula,For the actual measurement impedance of path k,For the meansigma methods of all path impedance, conversion parameter θ dimensionless, only relevant with alternative number of path, the excursion quite stable of θ is found through experiments, generally takes θ=3.3

2) the actual measurement impedance in the present invention, in formula (2)It is defined as the generalized cost V in section:

$c_k^{rs}=V={\mathrm{\β}}_1{l}_{1}a+{\mathrm{\β}}_2{l}_{2}b+{\mathrm{\β}}_3{l}_{3}c+{\mathrm{\β}}_4*1---\left(3\right)$

Then survey impedanceFor the linear function that parameter is β.

In formula, whether a is be highway (being that highway takes 1, be not that highway takes 0)；

Whether b is be major trunk roads (being that major trunk roads take 1, be not that major trunk roads take 0)；

Whether c is be branch road (being that branch road takes 1, be not that branch road takes 0)；

l₁、l₂、l₃Respectively corresponding road section length；

β₄* 1 for considering the number in section, the impact that namely expense is produced by the number of crossing.

3) by the actual measurement impedance redefined and formula (3), substitute in Logit modular form (2), then the probability of selecting paths kBe a unknown parameter it is the function of β.

$p_k^{rs}\left(\mathrm{\β}\right)=\frac{\exp \[-{\mathrm{\θc}}_k^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}{\underset{l\∈R_{rs}}{\Σ}\exp \[-{\mathrm{\θc}}_l^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}---\left(4\right)$

4) assume that the favored pathway that driver history selects is k₁, then the probability in path selected by him, i.e. selecting paths k is selected by formula (4) known driver₁Probability be:

$p_{k1}^{rs}\left(\mathrm{\β}\right)=\frac{\exp \[-{\mathrm{\θc}}_{k1}^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}{\underset{l\∈R_{rs}}{\Σ}\exp \[-{\mathrm{\θc}}_{k_1}^{rs}\left(\mathrm{\β}\right)/\stackrel{\‾}{c}\left(\mathrm{\β}\right)\]}---\left(5\right)$

5) assume that the driving historical data collected is 100 people, wherein, n₁Personal selection path k₁, n₂Personal selection path k₂.....n_nPersonal selection path k_n, wherein k₁, k2...k_nIt it is the set in all paths；Driver selecting paths k₁Probability be p (k₁), driver selecting paths k₂Probability be p (k₂) ... driver selecting paths k_nProbability be p (k_n)

The probability P (β) that then this sample occurs is represented by

P (β)=p (k₁ ⁿ¹·k₂ ⁿ²…k_n ⁿⁿ)

Take the logarithm, then

logp(k₁ ⁿ¹·k₂n2…k_n ⁿⁿ)=logpⁿ¹(k₁)+logpⁿ²(k₂)+...+logpⁿⁿ(k_n)

P (β)=n₁logp₁(β)+n₂logp₂(β)+...+n_nlogp_n(β)

Log-likelihood equation group is:

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_1}=0---\left(6\right)$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_2}=0---\left(7\right)$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_3}=0---\left(8\right)$

$\frac{\∂P\left(\mathrm{\β}\right)}{\∂{\mathrm{\β}}_4}=0---\left(9\right)$

β is estimated by formula (6), (7), (8), (9)₁、β₂、β₃、β₄。

6) according to the generalized cost V in the section demarcated and formula (3), the most frequently used shortest path is utilized to calculate heuristic algorithm--dijkstra's algorithm.

Dijkstra (Di Jiesitela) algorithm is proposed in nineteen fifty-nine by Holland computer scientist Dick Si Tela.Being be typically used for calculating the shortest path first from a summit to all the other each summits, solution is shortest route problem in directed graph.It is mainly characterized by centered by starting point outwards to extend layer by layer, until expanding to terminal.Dijkstra's algorithm can draw the optimal solution of shortest path.

Specifically comprising the following steps that of Dijkatra algorithm

Step0: initialize.Make i=0, s₀={ v_s, P{v_s}=0, λ (v_s)=0, rightMake T (v)=∞, λ (v)=M, make the label k=s of current check point；

Step1: terminate inspection.If s_i=N, algorithm terminates, now, c (v_s, v)=P (v),Otherwise proceed to step2；

Step2: amendment T label.(v is made to each_k, v_j) ∈ A andNode, if T (v_j) > P (v_k)+t_kj, then amendment node v_jT label, make T (v_j)=P (vk)+t_kj, λ (v_j)=k；Otherwise proceed to Step3；

Step3: determine P label.OrderIf T is (v_ji) <+∞, then v_jiT label be set to P label, i.e. P (v_ji)=T (v_ji), with S in season_i+1=S_i∪{v_ji, k=j_i.Then make i=i+1, proceed to Step1；Otherwise, algorithm terminates, now to each node v ∈ S_i, have c (v_s, v)=P (v), and to eachc(v_s, v)=T (v).

After the inventive method terminates, λ value " backward tracing " can be passed through and arrive from start node v_sTo the shortest path of any node v, namely find the preference that meets driver, the path that generalized cost is the shortest.

Above-described embodiment is only the preferred embodiment of the present invention; it is noted that, for those skilled in the art; under the premise without departing from the principles of the invention; some improvement and equivalent replacement can also be made; the claims in the present invention are improved and are equal to the technical scheme after replacing by these, each fall within protection scope of the present invention.

Claims (4)

1. the city path finding method considering driver individual's preference, it is characterised in that the method comprises the steps:

1) obtain driver trip historical data, described historical data be in the trip that driver is former select concrete path and the category of roads in path, length, by crossing quantity；

2) calculate the generalized cost V in section as follows, and it can be used as actual measurement impedance:

Wherein, k is certain paths selected, and r is starting point, and s is settled point,For the actual measurement impedance of path k between starting point r and settled some s；V is section generalized cost；Whether a is be highway: is that highway takes 1, is not that highway takes 0；Whether b is be major trunk roads: is that major trunk roads take 1, is not that major trunk roads take 0；Whether c is be branch road: is that branch road takes 1, is not that branch road takes 0；l₁For express highway section length, l₂For turnpike road segment length, l₃For branch road road section length；β₁For express highway section weight coefficient, β₂For major trunk roads section weight coefficient, β₃For branch road section weight coefficient, β₄For crossing weight coefficient；Wherein β₄* 1 for considering the number in section, the impact that namely expense is produced by the number of crossing；

Described β₁、β₂、β₃、β₄All try to achieve as follows: based on multinomial Logit mode, try to achieve the probability of selecting paths k；Based on historical data, estimate express highway section weight coefficient, major trunk roads section weight coefficient, branch road section weight coefficient, crossing weight coefficient with maximum likelihood estimate respectively；

3) utilize Dijkstra shortest path first, find the path that the generalized cost meeting drivers preference is the shortest.

2. consider the city path finding method of driver individual's preference according to claim 1, it is characterised in that described step 2) in, obtain the probability of selecting paths k in such a way:

It is primarily based on logit model, solves according to following formula and select the probability of path k between starting point r and settled some s

Wherein,For the actual measurement impedance of path k between starting point r and settled some s,For the meansigma methods of all path impedance, θ is conversion parameter, R_rsBeing the set in all paths between starting point s and settled some r, l is certain paths in set of paths；

Then the probability of selecting paths k is calculated according to following formula:

Wherein, β represents β₁、β₂、β₃Or β₄。

3. the city path finding method considering driver individual's preference according to claim 1 or claim 2, it is characterized in that, described step 2) in estimate express highway section weight coefficient, major trunk roads section weight coefficient, branch road section weight coefficient, crossing weight coefficient as follows:

Between the starting point r and the settled some s that are drawn by historical data, the probability P (β) of Path selection situation is:

P (β)=n₁logp₁(β)+n₂logp₂(β)+...+n_nlogp_n(β)

Wherein, β represents β₁、β₂、β₃Or β₄, n₁For the selected number of times in path 1, p₁(β) for the selected probability in path 1, n₂For the selected number of times in path 2, p₂(β) for the selected probability in path 2, n_nFor the selected number of times of path n, p_n(β) for the selected probability of path n；

Then β is estimated according to following log-likelihood equation group₁、β₂、β₃、β₄:

。

4. according to claim 1 or claim 2 consider driver individual's preference city path finding method, it is characterised in that described step 3) in Dijkstra shortest path first comprise the following steps:

Step0: initialize.

Step1: terminate inspection.

Step2: amendment T label.

Step3: determine P label.