WO2022126152A1 - Methods for estimating traffic loads and optimising road networks - Google Patents

Abstract

The present invention relates to a computer-implemented method for estimating traffic loads (L_jk) for different modes of transport (M_k) in a road network (RN) having nodes (N_i) and roads (R_j) the method comprising: selecting (S1 - S3) paths ( P_p) and for each path at least one mode combination (MC_m) and at least one trip (TRpmn); determining (S4 - S6) for each of said modes of transport an average physiological power consumption ( PC_k) for each of said roads and modes of transport, a physiological energy consumption (EC_jk) and for each of the mode combinations (MC_m) a physiological energy budget (EB_m); for each of the trips: computing (S7) a physiological energy effort (EE_pmn and computing (S8) a trip probability (TP_pmn) and estimating (S9) for at least one road (R₁) the traffic load (L_1k) for each mode of transport this road is configured for.

Description

Methods for Estimating Traffic Loads and Optimising Road Networks

The present invention relates to a computer- implemented method for estimating traffic loads for different modes of transport in a road network, the road network having roads and nodes between said nodes, each road being configured for at least one of said modes of transport. The invention further re- lates to a computer- implemented method for optimising a road network using said traffic estimation method.

The implications of human mobility can be found not only in transportation sciences and spatial economics, but also in social sciences and politics due to the manifold effects it has on the individual forms of mobility, infrastructure and urban planning, but most notably in respect to the environment, en- ergy and climate. In relation to this complexity, the most im- portant variables characterising and explaining mobility behav- iour are generally assumed to be with the entities of general- ised costs, i.e. travel time and - as part of the socio- economic variables - travel costs, car-ownership or income, culminating in the socio-economic approaches of utility theory, cf . , e.g. , H. Barbosa, et al. "Human mobility: Models and ap- plications", Physics Reports, Vol. 734, 2018, pp . 1 - 74. De- spite the basic nature of these measures and their extreme im- portance for mobility modelling, their functional descriptions and respective predictions are acknowledged to be ambiguous and have not been combined in a consistent model. Also, approaches explaining the extent of daily travel time itself have stated a law of constant travel time or a universal constant across space and time at around 1 to 1.3 hours per day. Although this phenomenon of the so-called travel time budget (TTB) has been recognised since the early 1960-ies, it is generally given as average statistics without decisive reasons for its stability or conclusive underlying functional relationships. Other concepts of travel behaviour modelling have often been made in analogies to the original physical concepts and especially Newton's mechanics, which have been used in two re- spects: Firstly, in terms of Newton's laws of motion as, e.g. in pedestrian modelling in D. Helbling, et al. , "Self- Organizing Pedestrian Movement", Environment and Planning B: Planning an Design, Vol. 28, 2001, pp . 361 - 383. SecoN_Dpy, in terms of the generally known gravity law, i.e. the Gravity Model (GM) , where the frequency of trips is proportional to the "masses of origin and destination" and the relative distance (function) , which has been evaluated in the early stages by ticket prices of train, bus and aeroplane. Further developments evaluated variants of the model structure with respect to com- muting or migration (F. Simini, et al. , "A universal model for mobility and migration patterns", Nature, Vol. 484, 2012, pp . 96 - 100) , opportunity and radiation modelling (E. Ruiter, "To- ward a better understanding of the intervening opportunities model", Transportation Research, Vol. 1, 1967, pp . 47 - 56) , comparisons of gravity vs. scaling, maximum entropy (A. Wilson, et al. , "Entropy in Urban and Regional Modelling: Retrospect and Prospect", Geographical Analysis, Vol. 42, 2010, pp . 364 - 394) , or energy (R. Kblbl , et al. , "Energy and Scaling Laws in Human Travel Behaviour", New Journal of Physics, Vol. 5, 2003, pp . 48.1 - 48.12) . Many of these studies have been developed based on motorised means of transport, foremost on car travel, and in combination with one specific travel purpose, i.e. com- muting. However, if we envisage a general model then all modes of transport, including the active modes with walking and cy- cling, and all purposes have to be equally accounted for, irre- spective of their current level of the modal split.

In general, methods for estimating traffic loads are widely used in traffic or area planning, road network construc- tion, maintenance and surveillance. Estimated traffic loads form the technical basis in determining parameters such as road condition, noise exposure, green house gas emission and road capacity utilisation. Therefore, traffic loads are invaluable to road network constructors and operators. In particular, an estimation of traffic loads with respect to different modes of transport such as walking, biking, car driving, taking public transport such as buses, metros or trams enables a mode- specific planning and constructing of the road network.

For example, an operator can dedicate (configure) roads with high estimated car traffic specifically to carpools or public transport to avoid traffic jams, estimate road abrasion and adjust maintenance intervals thereto to prevent pot-holes, estimate noise exposure to decide for locations of noise pro- tection walls, identify roads which could stimulate bike usage and dedicate them solely to bikes to reduce greenhouse gas emissions, et cetera.

The estimation of traffic loads is currently mainly car- ried out by methods relying on socio-economic variables, such as income, travel costs or car ownership. Socio-economic vari- ables, however, are difficult to obtain, e.g. from surveys, macro- or micro-economical data analysis. Moreover, once ob- tained it is difficult to build a consistent model from socio- economic variables due to their mutual interferences and corre- lations. Model inconsistency like a violation of basic physical principles inevitably impairs accuracy and is typically tackled by constructing more complex models, though, slowing down their computation. In addition, socio-economic variables are not uni- versally applicable as they change with the socio-economic en- vironment, e.g. , with economic wealth, mobility demands and Zeitgeist .

In alternative methods, traffic loads are estimated by pa- rameterised models stemming from physical considerations, e.g. , Newton's Laws of Motion or Gravity, etc. While these models ease model consistency, their parameters have to be fitted to observed data. However, models with a small number of fit pa- rameters suffer from inaccuracy and models with a high number fit of parameters suffer from their complexity which requires a lot of observed data and computing time, e.g. , to perform a high dimensional regression algorithm to fit the data.

Moreover, estimating traffic loads for different modes of transport requires the introduction of additional socio- economic variables or fitting parameters and, hence, worsens accuracy and the speed of computation even more.

It is, thus, an object of the present invention to over- come these limitations in the state of the art and to provide a computer- implemented method for estimating traffic loads for different modes of transport in a road network which provides an accurate estimation of traffic loads at a low computational effort .

This object is achieved with a computer- implemented method as mentioned at the outset, the method comprising: selecting, in said road network, paths formed by roads, each path connecting an origin and a destination node of the road network, and for each path at least one mode combination of the modes of transport the roads forming the respective path are configured for and at least one way of traversing this path with this mode combination as a trip, determining for each of said modes of transport an average physiological power consumption, for each of said roads, for each mode of transport this road is configured for, from a length of this road and the average physiological power consumption of this mode of transport a physiological energy con- sumption, and for each of the selected mode combinations a physio- logical energy budget in a time interval; for each of the selected trips : computing, from the physiological energy consumptions of the roads and modes of transport of this trip, a physiological energy effort of this trip, and computing, from said physiological energy effort and the physiological energy budget of the mode combina- tion of this trip, a trip probability of this trip; and estimating, for at least one road in the road network, from the trip probabilities of all trips whose path includes this road, the traffic load for each mode of transport this road is configured for.

The inventive method is based on the finding that traffic loads, even of different modes of transport, can be estimated from energies, i.e. , physiological power consumptions, energy consumptions and energy budgets, in an accurate and computa- tionally fast manner.

The accuracy of the method is owed to the finding that, irrespective of the socio-economic environment, travelling peo- ple are willing to "spend" on average a certain energy budget in a time interval for travelling with a given mode combina- tion, e.g. , 800 KJ/day for the "onefold-combination" of biking only, 1200 KJ/day for the "twofold-combination" of biking and car driving, 1600 KJ/day for the "threefold-combination" of walking, biking and car driving. Thus, by comparing the physio- logical energy effort of travelling a certain path using a cer- tain mode combination with the "available" physiological energy budget, an accurate trip probability is computed resulting in traffic loads accurately estimated therefrom. As only a com- parison of the physiological energy effort with the physiologi- cal energy budget is required to compute the trip probability, the method can be carried out in a fast manner without any ad- ditional fitting parameters and complex modelling.

The computational speed of the method is also owed to en- ergy being an additive quantity. Due to the additivity of en- ergy, the physiological energy consumption on each road with a specific mode of transport can be determined only once to form the basis of the physiological energy effort computation: One physiological energy consumption of a specific road can be used several times, i.e. , for computing the physiological energy ef- fort of all paths including this specific road. Thus, the en- ergy efforts of paths can be calculated particularly fast sim- ply by adding the energy consumptions of the respective path for the respective mode combination.

Last but not least, the physiological power consumption, energy consumption and energy budget are technical measures di- rectly accessible to physical measurement and independent of the social or economic environment, wherefore the method is universally applicable to road networks around the world.

The trip probability can either be a probability, e.g. , of one person traversing a certain path with a certain mode combi- nation, or a number indicating a probable absolute number of trips, e.g. , a hundred car drivers and bikers traversing said path with said mode combination. As a realisation of the former or a foundation of the latter, in a preferred embodiment the trip probability is calculated from an Erlang distribution of the physiological energy effort with its mean value being the physiological energy budget of the mode combination under con- sideration. An Erlang distribution guarantees that trips re- quiring a high physiological energy effort are exponentially damped out, which provides a particularly accurate modelling of trip probabilities of longer paths.

In this embodiment a particularly accurate modelling also of short trips can be reached when the shape parameter of the Erlang distribution is larger than two. Such a shape parameter results in a linear, quadratic, cubic, etc. energy effort term multiplied by the exponential and, thus, in a damping-out of short trips. This relies on applicant's finding that short trips while being energetically very favourable are not per- formed too often for the lack of accessible destination nodes within a short range around an origin node.

In a particularly favourable embodiment, the trip prob- ability is modelled by a linear energy effort multiplied by an exponential damping and computed according to

with

TP_pmn > being the trip probability of the trip under consid- eration ;

EE_pmn > being the computed physiological energy ef fort of the trip under consideration ; and

EB_m > being the physiological energy budget in said time interval for the respective m- th mode combination of the trip under consideration .

In general , one mode combination could be as likely as an- other . However , mode combinations including a larger number of modes of transport are typically less likely to occur . Hence , it is advantageous when trip probability of each trip is weighted with a factor depending on the respective mode combi - nation of its trip , to also account for an unequal distribution of mode combinations . This factor may, e . g . , be a global fac - tor , i . e . , be the same for the trip probabilities of each path, to obtain a global model .

In a favourable variant of this embodiment the factor is a local factor depending on the trip under consideration, wherein a participation value is assigned to each origin node for each selected mode combination, and wherein said factor also depends on the participation value of the selected mode combination of the origin node of the path of the trip under consideration . Thereby, the participation value accounts for local dif ferences in the distribution of mode combinations . For example , origin nodes with a younger population may have a smaller percentage of driver ' s licence holders and cars , and origin nodes in a ru- ral region may have less accessible public transport ; these origin nodes may then have a smaller assigned respective par- ticipation value for mode combinations including car driving or taking public transport , respectively .

In a further embodiment the trip probabilities of all trips with the same origin node are normalised so that their sum yields one. Normalised trip probabilities allow to obtain an estimation of the actual number of trips that will be car- ried out on a certain path per person.

To estimate absolute traffic loads, for example measured in cars, bikes, passengers per time interval on a specific road, in a favourable embodiment a population density is as- signed to each origin node and the trip probability of each trip is weighted by the population density assigned to the ori- gin node of the path of the trip under consideration. Of course, from absolute traffic loads a more significant estima- tion of road conditions, noise exposures, green house gas emis- sions and road capacity utilisations can be performed.

Alternatively or in addition thereto the population den- sity can fulfil a second purpose of prioritising paths by as- signing a population density to each node of the road network and selecting only those paths whose origin node has been as- signed a population density above a given density threshold. This allows to select only important paths with a high popula- tion density assigned to their origin nodes and on which a high traffic load can be expected. Neglecting the other (unimpor- tant) paths speeds up the estimation of the traffic loads with- out a loss in accuracy.

Similar to the population density of origin nodes, a des- tination attractivity can be assigned to each destination node and the trip probability of each trip can be weighted by the destination attractivity assigned to the destination node of the path of the trip under consideration. A higher destination attractivity could, e.g. , be assigned to nodes hosting shopping malls, sights, metro stations, train stations, airports etc. such that more accurate absolute traffic loads can be estimated therefrom .

Alternatively or in addition thereto the destination at- tractivity can also fulfil a second purpose of prioritising paths by assigning a destination attractivity to each node of the road network and selecting only those paths whose destina- tion node has been assigned a destination attractivity above a given attractivity threshold. Similar to population densities, destination attractivities allow to prioritise important paths with a high destination attractivity assigned to their destina- tion nodes over less important paths with a low destination at- tractivity assigned to their destination nodes. Again, consid- ering only important paths speeds up the estimation of the traffic loads without a loss in accuracy.

Especially for non-motorised modes of transport an eleva- tion profile of a road or path may be crucial for the choice of the mode of transport and the mode combination, respectively. In order to account for an additional physiological energy con- sumption on ascending roads and a decreased physiological en- ergy consumption on descending roads, in an advantageous em- bodiment a topology is assigned to each road and the physio- logical energy consumption is computed also from the topology assigned to the path under consideration.

As mentioned at the outset the estimated traffic loads are an important technical basis for road network constructors and operators. To present the estimated traffic loads in a particu- larly useful way to the user they can be visualized on a dis- play. The traffic loads could, for example, be visualised by different colours, e.g. , green, yellow, and red for low, inter- mediate, and high traffic loads, respectively, such that the user can immediately see critical red roads which have to be maintained more often, enlarged by building another lane, by- passed by new roads, etc.

In a second aspect the invention provides for a computer- implemented method for optimising a road network with respect to a predefined traffic target, comprising: a) selecting the road network as an input road network; b) performing the traffic load estimation method dislosed herein - in any of its embodiments or variants - to estimate for each road of the input road network a traffic load for each mode of transport this road is configured for; c) computing, from the estimated traffic loads of the in- put road network, a deviation of the input road network from the predefined traffic target; d) varying at least one of an arrangement of the nodes and roads of the input road network, the modes of transport the roads of the input road network are configured for, the as- signed participation values, if any, the assigned population densities, if any, and the assigned destination attractivities , if any, to obtain a candidate road network; e) performing the above-mentioned traffic load estimation method on the candidate road network to estimate for each road of the candidate road network a traffic load for each mode of transport this road is configured for; f) computing, from the estimated traffic loads of the can- didate road network, a deviation of the candidate road network from the predefined traffic target; and g) determining whether the deviation of the candidate road network is below a predetermined deviation threshold and, if so: selecting the candidate road network as the optimised road network and exiting the method; and h) determining whether the deviation of the candidate road network is smaller than the deviation of the input road network and, if so: selecting the candidate road network as new input road network; i) returning to step d) .

The optimisation method of the invention exploits the ad- vantages of the disclosed estimation method to design a new road network or improve an existing road network, e.g. , by building additional roads to homogenize traffic loads, or by enlarging roads to avoid traffic jams. Moreover, new nodes can be introduced, e.g. , to plan the location of a new city, air- port, shopping mall, and existing nodes can be modified, e.g. , by changing the assigned participation values, population den- sities or destination attractivities. Once an optimised road network has been found, in a pre- ferred embodiment of the invention this optimised road network is subsequently constructed on site (in "brick and mortar") .

The invention will now be described by means of exemplary embodiments thereof with reference to the enclosed drawings, in which show:

Fig. 1 a road network with nodes, roads therebetween and exemplary paths of respective trips in a schematic top view;

Fig. 2 a section of a road of the road network of Fig. 1 and respective traffic loads of different modes of transport on this road in a top view in detail.

Fig. 3 a method for estimating the traffic loads in the road network of Fig. 1, particularly on the road of Fig. 2, in a flow chart;

Fig. 4 experimental data of average power consumptions per length, per time and a respective average velocity for differ- ent modes of transport as used in the method of Fig. 3 in a ta- ble ;

Fig. 5 experimental data of participation values of dif- ferent mode combinations as used in an optional embodiment of the method of Fig. 3 in a probability/energy diagram;

Fig. 6 an exemplary Erlang distribution as used in the method of Fig. 3 in a probability density/energy graph; and

Fig. 7 a method for optimizing the road network of Fig. 1 by means of the method of Fig. 3 in a flow chart.

Fig. 1 shows a road network RN with nodes N_i (i = 1, 2, ..., I) and roads R_j (j = 1, 2, ..., J) . Each road R_j directly con- nects two neighbouring nodes N_i . Each node N_i of the road net- work RN can represent a traffic origin, road junction or traf- fic destination, e.g. , a single house or building, a residual area, a village, a city etc. In this sense the road network RN can be of any size and granularity, i.e. , a small sized road network of a village including also small roads, an intermedi- ate sized road network of a town, or a large sized road network of a region, country or state including only main roads. The road network RN can be an isolated road network or part of an- other road network.

Each road R_j of the road network RN is configured for and, hence, allows being traversed by at least one of different modes of transport M_k (k = 1, 2, ..., K) , as shown exemplarily in Fig. 2 for the road R₁ which is configured for the mode "walk- ing" M₁, the mode "biking" M₂ and the mode "car driving" M₃. Of course, this and other roads R_j can be configured for other modes of transport M_k, e.g. , including public transport modes such as taking a bus, a tram, a metro, a train, etc. Generally speaking, the road R_j can be any area allowing for traffic and, e.g. , be or comprise a footpath, a paved or unpaved street with or without sidewalks, a highway with one or more lanes, a rail- way or tramway track, etc.

On each road R_j every of its modes of transport M_k causes a respective traffic load L_jk, i.e. , a certain frequency or probability of traffic of this mode of transport M_k on the road R_j.

A computer- implemented method 1 for determining these traffic loads L_jk on at least one of the roads R_j (here: exem- plarily on the road R₁) shall now be explained with reference to Figs. 1 - 6.

As shown in the flow chart of Fig. 3, in a first step SI of the method 1, paths P_p formed by roads R_j are selected in the road network RN. Each path P_p connects an origin node N_Op and a destination node N_Dp (p = 1, 2, ..., P) , cf . the exemplary paths Pi, P₂ including the road R₁ in Fig. 1 connecting the same node Ni as origin nodes N₀₁, N₀₂ of the paths Pi, P₂ with differ- ent nodes N₄, N₅ as respective destination nodes N_D1, N_D2 of the paths Pi, P₂.

In principle, any set of possible paths P_p connecting two nodes N_Op, N_Dp in the road network RN could be selected, e.g. , all paths P_p including the road R₁, all paths P_p passing the road R₁ in a preferred direction of traffic D₁, D₂ (Fig. 2) . In certain embodiments, as detailed below, even paths P_p not in- eluding the road R₁ at all can be selected, e.g. , if traffic loads L_jk for more than one road R₁ should be estimated.

Two optional variants of a path selection are presented in Figs. 1 - 3 : In the first variant, a population density PD_i in- dicating the number of people at a specific node N_i is assigned to each node N_i in a step Sla. Then, the assigned population densities PD_i are each compared with a predefined density threshold TH_D in a step Sla' and only those nodes N_i whose as- signed population density PD₁ exceeds the density threshold TH_D are considered as origin nodes N_Op in the selecting step SI . Consequently, only those paths P_p are selected in step SI whose origin nodes N_Op have a population density PD_i exceeding the density threshold PD_i.

In the second variant, a destination attractivity DA_i in- dicating the importance of a specific node N_i as a destination is assigned to each node N_i in a step Sib. For example, nodes N_i hosting shopping malls, sights, airports, etc. can be as- signed a higher destination attractivity DA_i in step Sib. Then, the assigned destination attractivities DA_i are each compared with a predefined attractivity threshold TH_A in a step Sib' and only those nodes N_i whose assigned destination attractivity DA_i exceeds the attractivity threshold TH_A are considered as desti- nation nodes N_Dp in the selecting step SI. Consequently, only those paths P_p are selected in step SI whose destination nodes N_Op have a destination attractivity DA_i exceeding the attractiv- ity threshold TH_A.

In a further step S2 for each of the paths P_p selected in step SI at least one mode combination MC_m of modes of transport M_k for traversing the path P_p under consideration (i.e. of the currently considered path P_p) is selected. A mode combination MC_m can be any combination of modes of transport M_k which is possible on a certain path P_p, for example, a onefold combina- tion of using only one mode of transport M_k for the whole path P_p (MC₁ in Fig. 2) , a twofold combination of using two modes of transport M_k each for specific roads R_j of the path P_p (MC₂ and MC₃ in Fig. 2) or a manifold combination (not shown) .

Alternatively, the steps S2 and SI can be carried out in a different order, i.e. , the mode combinations MC_k are selected first and then only paths P_p on which the selected mode combi- nations MC_k are possible are selected in the step SI .

In a following loop LP1 over each path P_p and mode combi- nation MC_m of this path P_p selected in step S2 a step S3 is carried out selecting at least one respective trip TR_pmn (n = 1, ..., N) . Each trip TR_pmn corresponds to one possible way w_n of traversing the path P_p under consideration with the mode combi- nation MC_m under consideration (i.e. of the current cycle) . For instance, the path P₁ can be traversed in a first way Wi (first trip TR₁₂₁) with the mode combination MC₂ by first traversing the road R₁ biking M₂ and then the road R₃ walking M₁ or in a second way w₂ (second trip TR₁₂₂) with the mode combination MC₂ by first traversing the road R₃ walking M₂ and then the road R₃ biking M₂. Of course, the origin and destination nodes N_Op, N_Dp of the underlying path P_p are also the origin and destination nodes N_Op, N_Dp of the respective trip TR_pmn.

In a second loop LP2 - which may also be carried out be- fore any of the selecting steps SI - S3 - for each of the modes M_k of transport a step S4 is carried out determining a respec- tive (i.e. , for the mode of transport M_k under consideration) average physiological power consumption PC_k. The average power consumptions PC_k can, for instance, be determined from measure- ments, e.g. , as shown in the table of Fig. 4 for different modes of transport M_k (wk - walking, vo - biking, cd - car driving, cp - car pooling, bs - bus taking, rt - rail taking) . As can be seen in this table, the power consumptions PC_k can, e.g. , be determined per length 1 (second column) or per time t (third column) .

In a third loop LP3 for each of the roads R_j and their re- spective modes of transport M_k for which they are configured a step S5 is carried out determining a respective physiological energy consumption EC_jk- The physiological energy consumption EC_jk indicates the "energy cost" of traversing the road R_j under consideration with the mode of transport M_k under considera- tion. As such, the respective physiological energy consumption EC_jk is determined in step S5 from the average physiological power consumption PC_k of the mode of transport M_k under consid- eration determined in step S4 and a length LT_j (Fig. 1) of the road R_j under consideration. This may, e.g. , be done in two ways: Either by multiplying the average physiological power consumption PC_k given per length with the length LT_j, or by multiplying the average physiological power consumption PC_k given per time with a traversing time T_p determined from the length LT_j of the road R_j under consideration (Fig. 1) and an average velocity V_k of the mode of transport M_k under consid- eration, see the exemplarily provided average velocities Vk in the fourth column of the table of Fig. 4.

In an optional variant, in a step S5a a topology Tj indi- cating the steepness of the road R_j is assigned to each road R_j of the road network RN. In this variant, the respective physio- logical energy consumption EC_jk is determined in step S5 also from the topology Tj of the road R_j under consideration, e.g. , such that non-motorised modes of transport M_k have an enhanced physiological energy consumption EC_jk for ascending roads R_j and a reduced physiological energy consumption EC_jk for descending roads R_j .

In a fourth loop LP4 , for each of the mode combinations MC_k selected in step S2, a step S6 is carried out determining a respective physiological energy budget EB_m in a time interval ΔT. The physiological energy budget EB_m indicates how much en- ergy people are willing to spend in the time interval ΔT in which the traffic loads L_jk are estimated, e.g. , per day, week, year, etc. , for travelling with a specific mode combination MC_m. As can be seen in Fig. 5 this energy budget EB_k mainly de- pends on the number of modes of transport M_k, that are involved in a certain mode combination MC_k. For example, people are willing to spend on average approximately 800 KJ/day, 1200 KJ/day and 1600 KJ/day for onefold, twofold and threefold mode combinations MC_k (ellipses 2 , 3 and 4 in Fig . 5 ) , respectively . From these values or other measured data , e . g . , measured for each specific mode combination MC_m independent of its n- f oldness , the respective energy budget EB_k may be determined in step S6 . Of course , the loop L4 does not require the outcomes of the loops L1 - L3 and, hence , can be carried out before or after any of these . This applies to all loops and steps of the method 1 which are independent from outcomes of other loops and/or steps and, hence , can be carried out before or after the independent loops and/or steps , as known to the skilled person .

In a following loop LP5 for each of the trips TR_pmn se- lected in step S3 and loop LP1 the following steps S7 and S8 are carried out .

In the step S7 , from the physiological energy consumptions EC_jk of the roads R_j of the trip TR_pmn under consideration and the modes of transport M_k used on these roads R_j in this trip TR_pmn , a respective physiological energy ef fort EE_pmn of this trip TR_pmn is computed . The physiological energy ef fort EE_pmn can be computed by adding up the respective energy consumptions ECj_k . For example , for a first way of traversing the path P_T with the mode combination MC₂ by first traversing the road R₁ with a bike M₂ and then the road R₃ walking Mi , the respective energy consumptions EC₁₂ and EC₃₁ can be added up to obtain the energy ef fort EE₁₂₁ of the trip TR₁₂₁ .

In the next step S8 from the physiological energy ef fort EE_pmn of the trip TR_pmn under consideration computed in step S6 and the physiological energy budget EB_m of the mode combination MC_m of this trip TR_pmn determined in step S6 , a respective trip probability TP_pmn is calculated . It shall be noted that the trip probability TP_pmn can either indicate a probability of perform- ing the respective trip TR_pmn or a number of probable trips TR_pmn , as will be detailed later on . Finally, in a last loop LP7, for each mode of transport MR the road R₁ is configured for, in a step S9 the respective traffic load L_1k is estimated from the trip probabilities TP_pmn of all trips TR_pmn whose path P_p includes the road R₁. For exam- ple, the traffic load L₁₂ for biking M₂ on the road R₁ can be estimated by adding up the trip probabilities TP_pmn of trips TR_pmn which include the road R₁ and on which biking M₂ on the road R₁ has been selected in step S3. Optionally, from these trip probabilities TP_pmn only the ones whose trip TR_pmn traverses the road R₁ in one traffic direction D₁, D₂ can be added to es- timate a direction-dependent traffic load L₁₂.

In an optional variant, the last loop LP7 also runs over each road R_j and each mode of transport M_k the roads R_j are con- figured for, to estimate in each step S9 a respective traffic load L_jk for each road R_j under consideration and each mode of transport M_k under consideration from the trip probabilities TP_pmn of all trips TR_pmn whose path P_p includes the road R_j under consideration .

Coming back to the step S8 of computing the trip prob- abilities TP_pmn, in this step S8 the trip probabilities TP_pmn can, e.g. , be computed simply by dividing the energy budget EB_m by the respective energy effort EE_pmn to compute an absolute number of trips TR_pmn for a person in said time interval ΔT. With reference to Figs. 3 and 6 the trip probabilities TP_pmn can, however, also be computed as probabilities in step S8 from an Erlang distribution (curve 5 in Fig. 6) , with the mean value of the Erlang distribution corresponding to the physiological energy budged EB_m of the mode combination MC_m under considera- tion. , i.e. , according to

with

TP_pmn being the trip probability of the trip TR_pmn under consideration, i.e. on the p-th path P_p under consid- eration with the k-th mode combination under consid- eration in the n-th way of traversing the path P_p un- der consideration;

EE_pmn .... being the computed physiological energy effort of the trip TR_pmn under consideration, i.e. for traversing the p-th path P_p under consideration with the k-th mode combination under consideration in the n-th way under consideration;

EB_m .... being the physiological energy budget in said time interval ΔT for the m-th mode combination under con- sideration; and α .... being a shape parameter of the Erlang distribution.

The shape parameter α can, e.g. , be larger than two, or - as shown in Fig. 6 - be exactly two, resulting in a computation of the respective trip probability TP_pmn according to

As can be seen in Fig. 6 the distribtution according to equation (2) damps out trips with a high energy effort EE_pmn compared to the respective energy budget EB_m exponentially, and trips TR_pmn with a low energy effort EE_pm compared to the re- spective energy budget EB_m linearly.

Before the traffic loads L_jk are estimated in loop LP7, the trip probabilities TP_pmn can optionally be refined in an op- tional loop LP6 over all selected trips TR_pm, in which any of the following steps S8a and S8b or their combination is carried out .

In the step S8a the trip probabilities TP_pmn are normal- ised. Therefor, the respective trip probability TP_pmn under con- sideration can be multiplied by a factor Z which depends, e.g. , on the trip probabilities TP_pmn of all trips TR_pmn sharing a com- mon origin Node N_Op .

In a first variant, the trip probabilities TP_pmn can be normalised such that the sum of all trip probabilities TP_pmn for trips starting at the same origin node N_Op yields 1 (one) , e.g. , by

wherein the sum over p¹ , m¹ , n¹ is over all trips TP_p'm'n' sharing the origin node N_Op. with the trip TP_pmn whose trip prob- ability TP_pmn is normalised.

In a second variant, the trip probabilities TP_pmn can be normalised such that the sum of all trip probabilities TP_pmn for trips starting at the same origin node N_Op and carried out with the same mode combination MC_m yields 1 (one) , e.g. , by

wherein the sum over p¹ , n¹ is over all trips TP_p.mn. shar- ing the origin node N_Op. with the trip TP_pmn whose trip probabil- ity TP_pmn is normalised.

In a third variant, the trip probabilities TP_pmn can be normalised such that the mean value of all trip probabilities TP_pmn for trips TR_pmn starting at the same origin node N_Op and carried out with the same mode combination MC_m yields the en- ergy budget EB_m of this mode combination MC_m, e.g. , by

wherein the sum over p' , n' is over all trips TP_pW shar- ing the origin node N_Op. and the mode combination MC_m with the trip TP_pmn whose trip probability TP_pmn is normalised.

In the step S8b, which can be carried out before or after the step S8a, the respective trip probability TP_pmn is weighted, e.g. , accoring to one of the following three - optionally com- binable - variants:

In a first variant the respective trip probability TP_pmn is weighted in a step S8bl with a factor F_m depending on the re- spective mode combination MC_m. This factor F_m prioritises mode combinations MC_m over others. For example, taking a bus and driving a car is a rare mode combination MC_m and may, thus, have a smaller factor F_m than biking and taking a train which is a more common mode combination MC_m. The factor F_m may be a global factor F_m, i.e. , the same for all nodes N_i, or a local factor F_m(N_i) , i.e. , depending on the origin node N_Op of the trip TR_pmn whose trip probability TP_pmn is weighted.

In an optional version of this first variant the factor F_m can also account for local differences of the accessibility of different mode combinations MC_m in that in a step S8bl ' a par- ticipation value PVim is assigned to each origin node Ni for each selected mode combination MC_m. Then, the factor F_m can also depend on the participation value PV_im of the selected mode combination MC_m at the origin node N_Op of the trip TR_pmn of the trip probability TP_pmn under consideration. Fig. 5 shows a relation of participation values PV_im of 75:20:5 between one- fold (ellipse 2) , twofold (ellipse 3) and threefold (ellipse 4) mode combinations MC_m. Of course, other relations between par- ticipation values PVim, e.g. , independent of their n-foldness, could be assigned as well.

In the second and third variants of weighting, the above- mentioned population densities PD_i and density attractivites DA_i may be re-used in steps S8b2 and S8b3 as follows.

In the second variant, the trip probability TP_pmn of the trip TR_pmn under consideration may be weighted, e.g. , by multi- plying it with the population density PD₁ assigned to the ori- gin node N_Op of this trip TR_pmn- The so obtained trip probabili- ties TP_pmn then indicate a frequency of the respective trip TR_pmn in the said time interval ΔT. In embodiments without the step Sla, the respective population density Di can be assigned to each origin node N_Op in a dedicated step S8b2.

In the third variant, the trip probability TP_pmn of the trip TR_pmn under consideration may be weighted, e.g. , by multi- plying it with the destination attractivity DAi assigned to the origin node N_Op of this trip TR_pmn. In embodiments without the step Sib, the respective destination attractivity DA₁ can be assigned to each origin node N_Op in a dedicated step S8b3. In an optional last step S10 the estimated traffic loads L_jk can be visualised on a display. The visualisation can, for example, be in the colours green, yellow and red to indicate, respectively, a low, intermediate and high traffic load L_jk. However, any other way to visually discriminate higher from lower traffic loads L_jk may be used as known in the art.

Summing up, after carrying out the steps SI - S10 in the respective loops LP1 - LP7, for at least one road R_j in the road network RN, a respective traffic load L_jk has been esti- mated for each mode of transport M_m this road R_j is configured for .

The method 1 described so far can be employed in a method 6 to optimise the road network RN with respect to a predefined traffic target, as will now be explained with reference to Figs . 1 and 7.

In a first step Sa of this method 6 the road network RN is selected as an "input" road network RN_I .

In a second step Sb the above-mentioned method 1 - in any of its embodiments or variants - is carried out on the input road network RN_I to estimate the traffic loads L_jk on at least one road R_j of the input road network RN_I, as has been de- scribed above.

In a subsequent step Sc a deviation dev_I of the estimated traffic loads L_jk of the input road network RN_I from a prede- fined traffic target is computed. The predefined traffic target can, for instance, be a homogenous distribution of traffic loads L_jk over the road network RN, a reduction of certain modes of transport M_m, e.g. , to reduce CO₂ emissions, a maximal traffic load L_jk without traffic jams, etc. The deviation dev_I can be computed by defining a target function which quantifies the traffic target, e.g. , a homogeneity of traffic loads L_jk, and inputting the estimated or visualised traffic loads L_jk into this target function.

Then the input road network is RN_I is optimised in a loop LP_g comprising the following steps Sd - Sh ' . In a first step Sd of the loop LP_g the input road network RN_I is varied to obtain a "candidate" road network RN_C . To this end, at least one of an arrangement, e.g. , the positions of the nodes N_i or the courses of roads R_j (see, e.g. , the dashed new road R₄ added in Fig. 1) of the input road network RN_I and the modes of transport M_k the roads R_j are configured for, is var- ied. This can, e.g. , be done by introducing additional or re- moving existent modes of transport M_k a road R_j is configured for, or by changing one or more of the assigned values (if pre- sent) , i.e. , the participation values PV_im, the assigned popu- lation densities PD_i, the assigned destination attractivities DA_i, etc.

In a next step Se the method 1 is applied another time to estimate the traffic loads L_jk in the candidate road network RN_C, similar to step Sb.

Then, in step Sf a deviation dev_c of the estimated traffic loads L_jk of the candidate road network RN_C from the predefined traffic target is computed, similar to step Sc.

In the comparison step Sg the deviation dev_c computed in step Sf is compared with a deviation threshold TH_dev. If the de- viation dev_c is below the deviation threshold TH_dev (branch "y" of the comparison step Sg) , the method 6 continues with step Sg ' of selecting this candidate road network RN_C as the opti- mised road network RN_O, and the method 6 is finished.

If, however, the deviation dev_c of the candidate road net- work RN_C, is not below the deviation threshold TH_dev (branch "n" ) , then the method 6 continues with comparison step Sh.

In the comparison step Sh the deviation dev_c of the candi- date road network RN_C is compared with the deviation dev_I of the input road network RN_I . If the deviation dev_c of the candi- date road network RN_C is not smaller than the deviation dev_I of the input road network RN_I (branch "n" ) , the method 6 returns to the step Sd of varying the input road network RN_I to start a new cycle of the loop LP_g and test another candidate road net- work RN_C . If , however , the deviation dev_c of the candidate road net - work RN_C is smaller than the deviation dev_I of the input road network RN_I (branch "y" of the comparison step Sh) , the method 6 proceeds to an intermediate step Sh ' in which the candidate road network RN_C is selected as the new input road network RN_I , and its deviation dev_c is selected as the new input deviation dev_I , before returning to the first step Sd of the loop LP_g .

Once an optimised road network RN_O has been found in this way, the road network RN_O can subsequently be constructed on site .

The invention is not restricted to the specific embodi - ments disclosed herein, but encompasses all variants , modif ica- tions and combinations thereof that fall within the scope of the appended claims .

Claims

Claims :

1. A computer- implemented method for estimating traffic loads (L_jk) for different modes of transport (M_k) in a road net- work (RN) , the road network (RN) having nodes (N_i) and roads (R_j) between said nodes (N_i) , each road (R_j) being configured for at least one of said modes of transport (M_k) , the method comprising : selecting (SI, S2, S3) , in said road network (RN) , paths (P_p) formed by roads (R_j) , each path (P_p) connecting an origin and a destination node (N_Op, N_Dp) of the road network (RN) , and for each path (P_p) at least one mode combination (MC_m) of the modes of transport (M_k) the roads (R_j) forming the respective path (P_p) are configured for and at least one way (w_n) of trav- ersing this path (Pp) with this mode combination (MC_m) as a trip (TR_pmn) , determining (S4, S5, S6) for each of said modes of transport (M_k) an average physiological power consumption (PC_k) , for each of said roads (R_j) , for each mode of trans- port (M_k) this road (R_j) is configured for, from a length (LT_j) of this road (R_j) and the average physiological power consumption (PC_k) of this mode of transport (M_k) a physiological energy consumption (ECj_k) , and for each of the selected mode combinations (MC_m) a physiological energy budget (EB_m) in a time interval; for each of the selected trips (TR_pmn) : computing (S7) , from the physiological energy con- sumptions (EC_jk) of the roads (R_j) and modes of trans- port (M_k) of this trip (TR_pmn) , a physiological energy effort (EE_pmn) of this trip (TR_pmn) , and computing (S8) , from said physiological energy effort (EE_pmn) and the physiological energy budget (EB_m) of the mode combination (MC_m) of this trip (TR_pmn) , a trip probability (TP_pmn) of this trip (TR_pmn) ; and estimating (S9 ) , for at least one road (R₁ ) in the road network (RN) , from the trip probabilities (TP_pmn) of all trips (TR_pmn) whose path ( P_p) includes this road (R₁) , the traffic load (L_1k) for each mode of transport (M_k) this road (R₁) is conf igured for .

2 . The method according to claim 1 , wherein the trip probability (TP_pmn) is calculated from an Erlang distribution ( 5 ) of the physiological energy ef fort (EE_pmn) with its mean value being the physiological energy budget (EB_m) of the mode combination (MC_m) under consideration .

3 . The method according to claim 2 , wherein the shape parameter (a) of the Erlang distribution ( 5 ) is larger than two .

4 . The method according to any of claims 1 to 3 , wherein the trip probability (TP_pmn) is computed according to

with

TP_pmn being the trip probability of the trip (TR_pmn) under consideration ;

EE_pmn being the computed physiological energy ef fort (EE_pmn) of the trip (TR_pmn) under consideration ; and

EB_m > being the physiological energy budget (EB_m) in said time interval for the respective m- th mode combina- tion of the trip (TR_pmn) under consideration .

5 . The method according to any of claims 1 to 4 , wherein the trip probability (TP_pmn) of each trip (TR_pmn) is weighted (S8bl ) with a factor (F_m) depending on the respective mode com- bination (MCm) of its trip (TR_pmn) .

6 . The method according to claim 5 , characterised by as - signing (S8bl ' ) to each origin node (N_Op) for each selected mode combination (MC_m) a participation value (PV_im) , wherein said factor (F_m) also depends on the participation value (PV_im) of the selected mode combination (MC_m) of the origin node (N_Op) of the path (P_p) of the trip (TR_pmn) under consideration .

7 . The method according to any one of claims 1 to 6 , characterised by normalising (S8a) the trip probabilities (TP_pmn) of all trips (TR_pmn) with the same origin node (N_Op) so that their sum yields one .

8 . The method according to any one of claims 1 to 7 , characterised by assigning ( S8b2 ) a population density (PD₁) to each origin node (N_Op) , wherein the trip probability (TP_pmn) of each trip (TR_pmn) is weighted (S8b) by the population density (PD₁) assigned to the origin node (N_Op) of the path ( P_p) of the trip (TR_pmn) under con- sideration .

9 . The method according to any one of claims 1 to 8 , characterised by assigning ( Sla) a population density ( PD_i) to each node (N_i) of the road network (RN) , wherein only those paths ( P_p) are selected ( SI ) whose ori - gin node (N_i) has been assigned a population density ( PD_i) above a given density threshold (TH_D) .

10 . The method according to any one of claims 1 to 9 , characterised by assigning ( S8b3 ) a destination attractivity (DA_i) to each destination node (N_Dp) , wherein the trip probability (TP_pmn) of each trip (TR_pmn) is weighted (S7b) by the destination attractivity (DA₁) assigned to the destination node (N_Dp) of the path ( P_p) of the trip (TR_pmn) under consideration .

11 . The method according to any one of claims 1 to 10 , characterised by assigning ( Sib) a destination attractivity (DA_i) to each node (N_i) of the road network (RN) , wherein only those paths ( P_p) are selected ( S1 ) , whose destination node (N_Dp) has been assigned a destination attrac - tivity (DA_i) above a given attractivity threshold (TH_A) .

12 . The method according to any one of claims 1 to 11 , characterised by assigning ( S5a) a topology (T_j ) to each road (R_j ) , wherein the physiological energy consumption (ECj_k) is computed also from the topology (Tj) assigned to the road (R_j) under consideration.

13. The method according to any one of claims 1 to 12, characterised by visualising (S9) the estimated traffic loads (L_jk) on a display.

14. A computer- implemented method for optimising a road network (RN) with respect to a predefined traffic target, com- prising : a) selecting (Sa) the road network (RN) as an input road network (RN_I) ; b) performing (Sb) the method of any one of claims 1 to 12 to estimate for each road (R_j) of the input road network (RN_I) a traffic load (L_jk) for each mode of transport this road (R_j) is configured for; c) computing (Sc) , from the estimated traffic loads (L_jk) of the input road network (RN_I) , a deviation (dev_I) of the in- put road network (RN_I) from the predefined traffic target; d) varying (Sd) at least one of an arrangement of the nodes (N and roads (R_j) of the input road network (RNj) , the modes of transport (M_k) the roads (R_j) of the input road net- work (RN_I) are configured for, the assigned participation val- ues (PV_im) , if any, the assigned population densities (PD_i) , if any, and the assigned destination attractivities (DA_i) , if any, to obtain a candidate road network (RN_C) ; e) performing (Se) the method of any one of claims 1 to 12 to estimate for each road (R_j) of the candidate road network (RN_C) a traffic load (L_jk) for each mode of transport (M_k) this road (R_j) is configured for; f) computing (Sf) , from the estimated traffic loads (L_jk) of the candidate road network RN_C, a deviation (dev_c) of the candidate road network (RN_C) from the predefined traffic tar- get ; and g) determining (Sg) whether the deviation (dev_c) of the candidate road network (RN_C) is below a predetermined deviation 28 threshold (TH_dev) and, if so: selecting (Sg' ) the candidate road network (RN_C) as the optimised road network (RN_O) and exiting the method; and h) determining (Sh) whether the deviation (dev_c) of the candidate road network (RN_C) is smaller than the deviation (dev_I) of the input road network (RN_I) and, if so: selecting (Sh' ) the candidate road network as new input road network; i) returning to step d) .

15. The method according to claim 14, characterised by constructing the optimised road network (RN_O) .