WO2005088493A1

WO2005088493A1 - Method and system for optimising transport missions

Info

Publication number: WO2005088493A1
Application number: PCT/DE2005/000426
Authority: WO
Inventors: Ingo Morgenstern
Original assignee: Ingo Morgenstern
Priority date: 2004-03-10
Filing date: 2005-03-10
Publication date: 2005-09-22
Also published as: EP1741052A1; DE102004011604A1

Abstract

The invention relates to a method for optimising transport missions. According to said method, at least one location-dependent transport parameter is detected, and probabilities are determined for said parameter, providing the basis for the optimisation of the transport mission. The invention also relates to a system for optimising transport missions, said system comprising detection devices for detecting at least one location-dependent transport parameter, information devices that are embodied in such a way as to inform users about a possibility for optimising the transport mission, taking into account the at least one location-dependent transport parameter, and calculation devices that are embodied in such a way as to determine probabilities for the at least one location-dependent transport parameter, the information devices also being embodied in such a way as to inform users about a possibility for optimising the transport mission, taking into account the probabilities.

Description

Procedures and systems for the optimization of transport tasks

description

The invention relates to methods and systems for optimizing transport tasks, such as, for example, the transportation of goods by forwarding agents or the forward movement of conveyor belts in the manufacture of products. In particular, the invention relates to methods and systems for optimizing location- and time-dependent transport and also methods and systems for optimal utilization of traffic networks and for reducing costs in transport tasks and production line production.

At present, methods and systems for optimizing location and time-dependent transportation, such as the transportation of goods by forwarding agents, are known in practice, actual data being recorded and made available to users via communication devices. This can be used to avoid traffic jams, for example, if the information is received in good time. However, it often happens that the information is received too late, so that it is no longer possible to bypass an actually existing traffic jam. Another problem is that the information is passed on even if the traffic jam no longer exists, so that a bypass and thus a greater effort are then unnecessarily.

However, the information, such as traffic information, does not really solve the problem, even if it is transmitted correctly so that an actual traffic jam can be avoided. Bypassing a traffic jam can resolve it faster, but those who bypass the reported traffic jam often create a new traffic jam on the bypass route, so that the individual user of the traffic information is not really helped with the latter. He wants to avoid a traffic jam and ends up in the next traffic jam.

The known methods and systems therefore suffer from the fact that they do not bring about optimization, but instead generate problems and unnecessary effort themselves.

It is therefore the aim of the present invention to create methods and systems in order to optimize location and time-dependent transportation. This aim is achieved with a method according to claim 1 and a system according to claim 29. Accordingly, the invention provides a method for optimizing transportation tasks, wherein at least one location-dependent transportation parameter is recorded, wherein probabilities are determined for the at least one location-dependent transportation parameter that are used as a basis for the optimization of the transportation task.

It is preferably provided that, in addition to the probabilities determined for the transport parameters, the optimization of the transport task is taken as a basis. Alternatively or additionally, it can be provided that the probabilities of the at least one location-dependent transportation parameter are used as the basis for the transportation task in the form of an error bar, and / or that the probabilities or possibly error bars are continuously updated or improved with inclusion of the at least one location-dependent transportation parameter ,

A preferred development consists in that a temporal variation or dependency of the at least one location-dependent transport parameter is recorded, and for this purpose time-dependent probabilities are determined which are instead of the at least one location and time-dependent given transport parameter or in addition to the at least one location and time-dependent given transportation parameters of the transportation task.

Furthermore, it can preferably be provided that a probability distribution is determined from a plurality of probabilities, which is used as the basis for the transport task instead of the location-specific and time-dependent given transport parameter or in addition to the location-specific and time-dependent given transport parameter.

It is further preferred if the location-dependent transportation time is recorded as the at least one location-dependent transportation parameter.

A preferred embodiment of the invention consists in that the ascertained probabilities or, if applicable, the ascertained probability distribution are used as a basis for route optimization. This can be further developed by using time-dependent probabilities for the travel times, which are constantly adapted to the current traffic situations, with preference being given to adapting the time-dependent probabilities to the current traffic situation via satellite-based systems, and / or preferably adapting the time-dependent ones Probabilities of the current traffic situation contains a damping factor that prevents chaotic behavior. According to other developments of this embodiment, it is provided in particular that tellite-based data from road toll systems are used, and / or that inaccurate, noisy GPS data are used to create the time-dependent probabilities, and / or that a cost function is expanded for optimization by a risk term that reflects the time-dependent probability distributions for the travel times, and / or that to optimize the probability distributions for the total travel times Total costs are calculated.

According to the invention, provision can also be made to use the moments of the time-dependent probabilities for cycle times of production processes for optimization.

A further preferred embodiment includes that physical, stochastic, genetic algorithms and / or the savings algorithm are used for optimization. Alternatively or additionally, an analog optimization with probability distributions can be used.

It is further preferred if probability distributions are used instead of deterministic data.

Another development of the invention consists in that a bouncing method is used for optimization. Alternatively or additionally, it can be provided in the method according to the invention that for optimization Simulated Annealing, Threshold Accepting, Great Deluge (deluge), genetic algorithms, savings algorithms with or without subsequent bouncing are used. A preferred further development of these variants consists in that a bombshell method is used for the optimization, wherein in the context of the bombshell method in particular random coordinates are generated with a random generator, a radius around each coordinate point is determined or determined or generated , in particular generated by the random generator, and only the sections of the transport route or possibly the tour lying within these radii will be optimized with the bouncing method. This can be further developed by the random generator constantly changing new random coordinates and / or radii

Random coordinates are generated.

It is further preferred if a risk assessment is included, the risk assessment in particular being carried out using the "value-at-risk" method.

In a further development of the invention, the method for optimizing transportation tasks can also provide for the Search Space Smoothing method to be optimized with or without threshold and / or low-temperature simulated annealing processes. Alternatively or additionally, a small threshold can be used when using threshold acceptance or a threshold method.

According to the invention, the stated goal is also achieved with a system for optimizing transport tasks, wherein detection devices are provided for recording at least one location-dependent transport parameter and information devices which are designed to provide users with inclusion of the at least one location-dependent transport parameter via an optimization possibility of the transport task inform, and furthermore calculation devices are provided which are designed to determine probabilities for the at least one location-dependent transport parameter, and wherein the information devices are designed to inform users, including the probabilities, of an optimization possibility of the transport task.

The refinements, developments and variants of the method according to the above information and explanations also apply in an analogous manner to the system for optimizing transport tasks according to the invention, in that devices for carrying out the method for optimizing transport tasks in accordance with these refinements, refinements and Variants included and can be operated as intended.

Process and device features also result from device or. Process descriptions. Unless otherwise specified here, the method features given give the person skilled in the art the suitable and therefore intended devices and devices without any special details, so that he does not have to be inventive in order to select or define them if he is aware of the disclosure of the present application ,

Preferred and / or advantageous embodiments of the invention result from the respective dependent claims and their combinations as well as the entire application documents.

The invention is explained in more detail below using exemplary embodiments with reference to the drawings, in which

1 shows by way of example an explanatory and illustrative traffic flow forecast for Munich at 10:00 p.m. 2 shows by way of example an explanatory and illustrative traffic flow forecast for Munich at 9:00 a.m.

3 shows an example of a probability distribution for the travel time of a route at a specific time,

4 shows, by way of example, a probability distribution for the travel time of the route according to FIG. 3 at a specific time, but different with respect to the representation in FIG. 3,

5 shows an example of a probability distribution for total costs,

6 shows an illustrative explanation of the “value-at-risk” method,

FFiigg .. 77 shows an illustrative explanation of the gradient method,

8 shows an example of a tour in a mathematical representation,

9 shows an example of a risk assessment using error bars,

10 shows an example of an overall distribution for calculating VaR,

11 shows an example of a risk component calculated as a value-at-risk in the total cost function,

12 shows an exemplary explanatory sketch for simulated annealing,

13 shows a schematic representation to represent the installation of various components at different stations of an assembly line, and

14 shows, by way of example and schematically, the overall probability for the installation of all parts in a predetermined order of the vehicles.

Using the exemplary embodiments and application examples described below and shown in the drawings, the invention is merely explained in more detail by way of example, i.e. it is not restricted to these exemplary embodiments and application examples or to the respective combinations of features within individual exemplary embodiments and application examples. Process and device features also result analogously from device or process descriptions.

Individual features that specify and / or are shown in connection with specific exemplary embodiments are not limited to these exemplary embodiments or the combination with the other features of these exemplary embodiments, but can, within the scope of what is technically possible, with any other variants, even if they are shown in FIGS available documents are not treated separately, can be combined.

Identical reference symbols in the individual figures and illustrations of the drawings denote identical or similar or identically or similarly acting components. On the basis of the representations in the drawing, such features are also made clear that are not Chen are provided, regardless of whether such features are described below or not. On the other hand, features which are contained in the present description but are not visible or shown in the drawing are also readily understandable for a person skilled in the art.

First, basic principles of methods and systems according to the invention for optimizing transport tasks are explained in more detail using an exemplary embodiment with reference to route optimization, for example of forwarding agents.

For the route optimization, time-dependent probability distributions are used in the exemplary embodiment dealt with here, which are constantly adapted to the current traffic situation with the aid of, in particular, satellite-based systems.

The avoidance of traffic congestion or even a traffic collapse is the goal of numerous efforts. In this context, the new toll systems such as "Toll

Collect "to mention in Germany. The basic idea of such toll systems is to influence the traffic by means of a time-dependent toll for trucks. The problem lies in setting the time-dependent toll, since exact regulation is only possible in" trial and error "procedures The system regulated in this way can therefore tend to be chaotic.

The exemplary embodiment of the method and system for optimizing transport tasks, which is considered in more detail here, circumvents this difficulty. The adjustment takes place automatically. The procedure in the present exemplary embodiment is as follows: The traffic data collected by the trucks, for example coordinates and times in satellite-based

Systems are collected centrally. On an underlying digital map, time-dependent probability distributions for the travel times are calculated for the road sections from the coordinates and associated times. These time-dependent probability distributions are made available to every road user, if possible even free of charge. B. on the Internet or in real time via suitable information facilities in the vehicles. The distributions mentioned are constantly updated and, if necessary, reprocessed in the event of new external influences (e.g. new construction sites etc.). This gives the user a probability statement for the Regensburg - Munich route, departure Friday, 9:00 a.m., etc. Observing these probability statements is now an important part of the overall system. Avoiding routes with a high probability of longer travel times at certain times leads to a shift in traffic jams. By updating the distributions with new ones Satellite data is a kind of feedback of the system or an actual-target comparison. In this way, the system regulates itself and approaches the optimum, ie the system and thus the process is capable of learning or self-learning. This regulation is the fundamental advantage of the new procedure. A damping factor can prevent tendencies towards chaotic behavior or vibrational behavior. The damping factor determines the degree to which the old satellite data are replaced by new ones.

To implement this system and method, it is sufficient that only some of the trucks are equipped with measuring systems. In addition, the measurement accuracy is of less importance than in legally usable data structures such as toll systems. This means that the current GPS data can be used with the known accuracy because, as mentioned, the creation of probabilities has a certain fault tolerance.

As mentioned above, an essential component of this exemplary embodiment of the system and method for optimizing transport tasks is the influence of the distributions on the driving behavior of the users. Based on the assumption that truck traffic is an essential factor in the flow of traffic, truck tour planning is of the utmost importance.

The previous route planning and optimization relates to purely deterministic data, e.g. B. from A to B exactly 60 minutes. It is therefore not possible to include traffic congestion. The method and system for optimizing transport tasks in accordance with the exemplary embodiment concerned in the present case includes route optimization which is based on the time-dependent probability distributions for the travel times from A to B mentioned. The distribution for an entire tour can be calculated for this in physical, stochastic or genetic algorithms. This procedure is time-consuming and must be replaced in the case of short response times or short-term required responses by treating the moments of the distribution. The simplest method is the use of the average travel time and the standard deviation, i. H. the error bar. On this

At the level, "simpler" optimization methods such as the savings algorithm can also be used. The end result in the optimization is no longer just the optimal travel time or the optimal costs, but an average travel time or average costs and a risk assessment z. B. with a 10% probability of an extension of the journey time by 30 minutes etc. This allows the user (forwarder or driver) to choose between different risk scenarios. This roughly corresponds to the risk assessment for portfolio creation on the stock market. Result of the introduction of this true Probability-based route planning is that trucks can avoid jam-prone, ie risky routes at the appropriate times and will benefit everyone involved. The system described is thus conclusive in itself. After a settling phase it will bring about the best possible utilization of the transport network.

Of course, the system can also be implemented on other modes of transport or their network.

A generalization of the exemplary embodiment dealt with above consists in the application of the probability-oriented optimization to production processes. A sub-process no longer takes 10 minutes deterministically, but the corresponding probability distribution becomes the basis of the optimization. This results in an application of the new method to almost all cases of deterministic optimization.

A further exemplary embodiment is now discussed, which relates to a method and system for optimizing transport tasks with a learning ability. With this adaptable procedure and system for the optimization of transport tasks, the optimal utilization of a traffic network, for example, can be achieved. This new method and system is based on inventive findings in algorithm generation, which make it possible to include time-dependent probabilities in route optimization. These time-dependent probabilities are based on measurements of the traffic flow. The ability to learn is based on a target-actual comparison with the measured data. Truck traffic is influenced in such a way that the main reasons for traffic jams are eliminated. The overall system is of great economic benefit, which will only increase if the most extensive possible introduction is achieved. For individual users, such as freight forwarding companies, there is an improvement in the business situation even with isolated introduction, since they can adapt to the traffic situation optimally. Some of the required traffic data is already available.

For an ordinary road user who only wants to get from A to B, with his

Route planning can be relevant to two criteria in particular: the expected journey time and the travel costs. By balancing these two factors, each weighted according to his personal needs, he will ultimately make the decision for a specific mode of transport and a specific route.

A freight forwarding company that has to send its trucks to certain destinations in order to fulfill customer orders and wants to plan the trips of its trucks for this purpose, faces a much more difficult problem. As far as the loading capacities of the trucks allow, several destinations should be approached in one trip for cost reasons, ie the trucks are usually on longer trips. In addition, the company can provide additional tours for a truck after returning from the first tour. The more you want to reduce downtimes at the fleet location, for example in order to achieve a good utilization of the vehicles, the more important it is to calculate the time required for each individual tour, which is made up of the travel times and length of stay at the customer. This scheduling becomes even more critical when you consider that most jobs need to be completed within one or more time slots.

The freight forwarding company is therefore faced with the task of distributing the individual customer orders in a suitable manner to the trucks in their own fleet and combining them into tours, and ensuring that the destinations are in the best order for each tour. So it has to draw up an extensive tour schedule for the near future. In practice, other restrictions are sometimes to be observed, for example because not every truck is suitable for every job.

Of course, every forwarding company is careful to avoid unnecessary costs. The prerequisite for an overall inexpensive tour plan, which fulfills all orders and satisfies all ancillary conditions, is also in this case knowledge of the travel times and travel costs of all trips in question. However, due to the complexity of this problem, you have to rely on uncomplicated and quick access to all relevant data (vehicle properties, order data, travel time and travel costs, etc.) when creating a tour plan, which is provided by the use of special tour planning software can. If you want to leave the search for the best solution entirely to the computer, you have to rely on heuristic optimization methods, because exact methods are overwhelmed due to the complexity.

Time dependency of the route data

The travel time for any journey from A to B can only be calculated as a first approximation from the route length, the type of road (highway, federal road, country road, etc.), the (legally permitted) speed of the vehicle and other factors that are constant over time. Your own progress in traffic is often hampered by the presence of other road users. If a particularly large number of road users are on the road, there may even be a standstill, i.e. a traffic jam. As a result, the travel time can be predicted more precisely if one includes data on the expected traffic volume includes his planning. In heavy traffic, so there is a risk of traffic jams, the average speed of the vehicles will probably have to be corrected slightly downwards. With very little traffic, ie with "free travel", you can calculate with the (permissible) maximum speed of the vehicle. Since the volume of traffic fluctuates greatly during the day, you have to deal with time-dependent travel times. Furthermore, the volume of traffic on individual sections of the route can be very different. It is therefore possible that, depending on the time of day, very different routes are optimal for driving from A to B.

1 and 2 show traffic flow forecasts for Munich at 10 p.m. and 9 a.m. only as illustrative examples. To this extent, this representation is self-explanatory and understandable for any person skilled in the art in connection with the present description.

The travel costs are initially the same as the travel times - at least if you assume operating costs per time. However, from the introduction of a toll for the journey on some routes (sections) one has to calculate additional costs, which are proportional to the route length and depending on vehicle characteristics. There are considerations to expand the toll system, for the introduction of which primarily financial reasons are given, so that traffic control is also possible via it. For example, one could abolish the Sunday driving ban and charge increased tolls to compensate for these days. It is also conceivable to control the traffic according to the time of day by trying to relieve the highways in favor of car traffic at higher times and at the same time create an incentive for driving at night with low fees. Perhaps this could lead to more balanced utilization of road capacities.

risk

Because forwarding companies sometimes have to make appointments, the desired minimization of travel times and travel costs is often accompanied by the desire to be able to meet the arrival time agreed with the customer with the greatest possible security. In the case of such forward transactions, it can in no way be assumed that in the long run, late and early arrival times will balance each other out in the cost balance sheet, because the two cases are to be valued completely differently in terms of cost. A late arrival at the customer can result in high follow-up costs for the forwarding company. On the other hand, if the truck arrives too early, there can be no benefit if the truck only has to wait a little longer for it to be processed. In such cases, it poses a great risk, for example on Friday afternoon between 1:00 p.m. and 3:00 p.m. to drive from Regensburg to Munich due to the slow traffic and only to rely on the predicted average travel time. In this context, you have to consider that the traffic models can often only derive a probability of a traffic jam from the forecast traffic volumes - so if you are lucky, you can drive through quickly, you are unlucky, if you get stuck in a traffic jam.

3 and 4, probability distributions for the travel time of the same route at different times are shown only as illustrative examples. Outside the peak times, for which an example is shown in FIG. 3, the average travel time is lower and the scatter is smaller, as the example of FIG. 4 illustrates.

If no important information is to be lost, the travel time for each road connection should therefore be given in the form of a time-dependent distribution. Such distributions are obtained by statistical analysis of measured traffic data. The greater the fluctuations in the measured travel times, the broader the distribution for the travel time. The average of this distribution is the average journey time. Their width can be viewed as a measure of the blurring or inaccuracy of the travel time and should also be taken into account when planning the tour.

Based on this probability information, you can draw up a tour plan where the risk of not being able to meet appointments is low, e.g. For example, by moving planned departure times for tours accordingly, changing the order of destinations within tours, or choosing different routes for some trips. The shipping company can decide to what extent it wants to reduce the risk and accept higher costs.

Feedback - adaptive system / procedure

The use of the described methods will result in the truck avoiding roads with a high probability of traffic jams. As a result, traffic will shift to less stressed regions and times. This can lead to new traffic jams. However, these new situations will be taken into account in the next tour planning. This again results in a shift in the traffic situation, which in turn leads to a changed route planning. The system thus reacts to both improvements and deteriorations in the overall situation. This results in an iterative approximation to the optimal use of the transport network. In order to achieve the broadest possible application of the invention, not only an independent tour planning program for implementing the method according to the invention and as part of a system according to the invention for optimizing transport tasks can be brought onto the market, which competes with the established programs, but can Invention can also be incorporated into already established route planning programs, systems and procedures.

In particular, it is possible to include traffic forecasts that have already been determined in the invention. A connection to existing digital map material is also possible. Furthermore, existing, in particular comprehensive, reliable recordings of traffic flow information can be included. Such data and information, regardless of whether they already exist or are yet to be obtained, are preferably from several different sources, such as. B. Traffic counting points (for example, nationwide or Europe-wide), traffic forecasts, digital maps, road construction planning, etc.

A further exemplary embodiment is described in more detail below.

For example, truck traffic on highways in Germany is used. However, this assumed exemplary situation can also be generalized for all roads and also for only or with the inclusion of cars.

Furthermore, in this example it is assumed that there is an overall system capable of learning, with data being collected, for example, by a toll system, such as, for. B. Toll-Collect in Germany, on the basis of a generally satellite-based system, as is customary at the current time. GPS is used here, the basics of which are assumed to be generally known and therefore are no longer explained separately here, and data is collected in a central computer about the time when a certain truck drives on a certain section of the motorway. The travel time for the section is then calculated from this. A histogram is created for different times of the day, days of the week, different seasons, etc. In turn, these histograms are used to create probability distributions, which are entered into a traffic forecast program and compared with reality. From this one obtains probability information for the "truth content" and thus a probability distribution. For example, route plans can be created based on such probability distributions.

Route planning - definition A tour planning exists if a forwarding agency specifies driving parameters with only one order (only one truck), up to the case that a forwarding agency with a certain number of trucks that have a certain number of orders have to carry out the entire fleet and all orders in plans, organizes and / or manages in an optimizing manner.

Multiple orders in succession are possible. There are also secondary conditions that are useful for the industry (time window, working day, etc.). In any case, tours therefore include a single truck that travels from Frankfurt to Munich, for example, but also a complex fleet, such as for fuel oil export.

The inventive step in the present exemplary embodiment again consists in the fact that instead of a deterministic statement about a travel time from A to B, a probability statement is made according to the invention as before. Optimization methods are also implemented, such as physical methods, e.g. Simulated annealing, threshold acceptance, Great Deluge (deluge), genetic algorithms, savings algorithms with or without subsequent bouncing ("bouncing" method) (one takes the solution, for example, of the savings algorithm as a starting configuration for the bouncing method, which later is described in more detail). Savings algorithms alone provide rather inadequate solutions, so that an improvement with bouncing based on a physical algorithm (simulated annealing, etc.) is preferably provided and achieved, which means optimization with probabilities.

With such a route planning or optimization, the result is a probability distribution for the total costs, for which an illustrative example is shown in FIG. 5. For the total costs, a risk assessment is preferably carried out with

"Value-at-Risk" procedure ("Value-at-Risk": VaR), d. H. corresponding to the area under the curve, as the illustrative illustration of FIG. 6 illustrates. 6 means "VaR 99": with 1% probability these costs, analogously the statement "VaR 95": with 5% probability would mean these costs, etc. This enables a selection of the subjective best tours Specification of how willing to take risks, conventional, etc., such as B. with equity portfolios.

Basics of learning ability are explained below. It is assumed that the system will be used across the board, for example by a large number of freight forwarders. The reality is that today trucks drive tours for given or predicted traffic situations. This means influencing the existing traffic situations, in particular in such a way that previous traffic jams disappear. One consequence of this is that new traffic jams occur in other places and at different times. There is therefore an interaction between truck tour planning and traffic situations. This is based on this and this independently optimizes the learning ability of the overall system. Always new and current GPS data automatically lead to a change in the probabilities, which in turn enables new tour plans by creating changed tour plans according to new traffic situations. The control mechanisms originate from the constant comparison of the probabilities and probability distributions with reality and include a constant "readjustment" through new tour plans. It is e.g. B. move around a gradient, in particular in the case of total traffic or only trucks, as is illustrated schematically by the illustration in FIG. 7. The system gradually approaches an overall optimum.

In a development of the exemplary embodiment dealt with here, further improvements to the system can be achieved, for example, by integrating one or a combination with a time-dependent toll. For example, routes on which traffic jams occur continuously or at certain times can be subject to an additional traffic-related toll depending on the time.

Such situations can be calculated using the probability distributions created as above. The relationship between traffic density on the one hand and travel times or speeds on the other hand is used, as is already known as a procedure. In the context of the invention, it was recognized that it may be best to calculate the toll based on the traffic density (simplest proposal: linear relationship). Other design options are to change the (time-dependent) toll depending on the effect achieved, such as: B. Gradually increasing slowly, etc. Furthermore, the system and method according to the invention enables a evasion of predicted problem cases (traffic jams, etc., high traffic volume, etc.), that is to say a greater deviation from the normal case. The toll can preferably be designed in such a way that trucks avoid these vulnerable points. Control can take place, for example, by e.g. B. lower toll if a certain section is not traveled at a certain time, or a higher toll for vulnerable areas. The greatest advantage of the method and system according to the invention for optimizing transport tasks is learning ability. This is very much supported when trucks are based on predicted probabilities.

Orientation to online data (real-time data) as they are currently used, which is therefore to be regarded as state of the art, does not lead to a learning effect. Nevertheless, such online data (real-time data) can be integrated into the system and method according to the invention and contribute to a further improvement. The aspect of the system and method according to the invention specifies probabilities and the online data indicate the reality. So z. For example, it can also be taken into account that a traffic jam is actually at one point, although only a low probability of traffic jam was determined for this point.

There now follow general considerations and explanations of the method and system for optimizing transport tasks as a whole, as well as individual aspects, as far as this is conducive to understanding, with reference to route planning problems for freight forwarders.

Truck forwarding agencies in Germany are essentially equipped with route planning systems. Almost all of these are based on savings algorithms, generally at different levels. For example, this is done as follows: First, customers are distributed to trucks. Then the overall situation is optimized. This is followed by a partial splitting off of constraints, theoretically according to the principle: "Devide and loose" (first customers on a truck and then planning a tour). This results in a deviation from the total optimum of z. B. five percent depending on the complexity of the tour planning.

The invention is replaced by or supplemented with new algorithms. In particular, there is basically an improvement of the previous route planning through subsequent "bouncing" processes. The consideration of the probability distributions only brings additional improvement in "bouncing", i.e. Savings solution. In other

Words this means that you add probabilities, do the bouncing and get the final solution according to the invention.

Explanatory descriptions of exemplary embodiments in the context of optimization processes now follow: As previously explained, the probability distributions are preferably taken into account as a function of time. This time dependency naturally also applies to the mean values. This enables a simplified procedure to be obtained: Instead of entire probability distributions, only error bars are used / taken into account, e.g. B. in simulated annealing or analogously in other methods. An exemplary explanatory mathematical representation thereof as a representation of a tour is shown in FIG. 8. Significant progress compared to pure simulated annealing is that the time dependency and the error bars are taken into account.

The essential simplification of this procedure consists in the additional term in the cost function of error bars; risk assessment is now carried out by the error bar of the overall distribution and not by VaR, but is equivalent assuming a Gaussian distribution. As previously explained, the probability distributions are time-dependent, so of course the mean values and the error bars. Proportionality factors therefore exist between the previous cost functions Ko and the proportion of error bars F.

As mentioned, the risk assessment is now done with error bars instead of value-at-risk. This is illustrated by the graphic in FIG. 9. The approach is correct for a Gaussian distribution and analogous to, for example, the Markowitz theory for

Equity portfolio. Proportionality factors γ (gamma) result in a risk-friendliness etc. analogous to Markowitz. Accordingly, the following applies to shares: In the event of large deviations from the Gaussian distribution, no longer count on error bars, but take along the entire probability distribution (standard procedure); i.e. no longer just adding up the error bars, but folding the probabilities (more time-consuming). This is illustrated by the representations of FIG. 10: Instead of error bars, now the overall distribution and the calculation of VaR from it. This in turn gives the overall distribution, as can be seen from the illustration in FIG. 11. The risk component in the total cost function is thus calculated as value-at-risk.

The bouncing process is explained below.

It is the start configuration: a few% of the optimum, ie the savings solution, as used in the existing widespread tour planning, is very suitable. Approachally, a warm-up takes place at maximum specific heat in simulated annealing or threshold etc. With reference to the explanatory FIG. 12, T stands for the virtual temperature temperatures in simulated annealing or threshold, as is common in physical optimization processes.

Simulation methods in physics

After warming up to Tf, a relatively rapid cooling to T against 0 is then repeated several times, i.e. H. bouncing. The repeated repetition (bouncing) takes place until the desired result is achieved.

A subspecies of the bouncing process is the so-called bombshell process. Instead of

Bouncing e.g. For example, to carry out an entire map, "bomb funnels" are set: only what is in it is bounced. That is, e.g. B. generates a random generator random coordinates and a radius around each such coordinate point, d. H. so-called "bomb funnels". Only the sections of the tour that lie within these radii are optimized with the bouncing method. In other words, a change in the tour in optimization processes only takes place within these radii, and thus bouncing, as indicated, only within the funnel. Random generator funnels are always being generated until the desired result is achieved, especially until no further improvements can be obtained.

Another suitable optimization method is search space smoothing.

Reference is now made to an improved search space smoothing as described in the literature, but instead of a greedy algorithm for local relaxation, a threshold method with a small threshold or simulated annealing with very low temperatures is used. Greedy is simulated annealing with T = 0 (only improvements). Thus a tour can be created with Search Space Smoothing with Threshold. The introduction of a small threshold or very low temperatures leads to considerable improvements in tour planning. This innovation, ie the use of threshold or low-temperature simulated annealing leads to the best results in tour planning. This procedure can also be used for bouncing. The smoothing parameter is changed in the same way as the temperature, ie switching as in bouncing of the improved search space smoothing, and thus the same procedure as in bouncing, only with changes in the smoothing parameters: ever greater change, analogous to bouncing; Combination with bombshell processes means only changing the funnels. An essential aspect of the exemplary embodiments dealt with above and the invention in this context is the inclusion of the error bars or the probability distribution and / or the time dependence in optimization methods. Further advantageous and preferred configurations relate to bouncing, bombshell and improved search space smoothing.

Now there is a discussion of configurations of the invention in the context of production lines. Fundamentally, a mathematical mapping of transport tasks in production lines to route planning problems is possible, i.e. an analogous approach. Algorithms can be used as described in connection with tour planning, whereby there are also improvements to previous results through bouncing and probabilities as well as possibly time dependency (e.g. shift changes, mental / physical conditions of the employees through festival time, regional and seasonal living conditions, ...) etc.

The sequence of vehicles on an assembly line is used for an exemplary embodiment. A wide variety of components are installed at different stations, as the schematic illustration in FIG. 13 illustrates. For example, a type j rear-view mirror is installed at installation location i. For the average installation time, a cantilever or a probability distribution is obtained. H. you get probability distributions for all installation locations and installation types: distribution for each individual type; Type for everyone

Vehicle specified. At a given time, certain vehicles are on the belt in certain places. There is a probability that a last component will be installed in time. Then the belt moves wide and new situations arise. This in turn includes probability distribution until the last component is installed. This results in total cost functions. The overall probability for the installation of all parts in a predetermined order of the vehicles is shown schematically in Fig. 14 for clarification, i.e. the average total installation time with risk. This risk follows from value-at-risk distributions, or from the error bars, as described above for trucks, again including the factor γ (gamma).

Usual changes to the configuration (order) can be based on known methods (simulated annealing, etc., but also genetic algorithms).

According to the invention, the introduction of the probabilities and in particular the time-dependent probabilities is also advantageous here. This results in a further advantageous design option when operating production belts: The belt does not only move further by one station if the longest assembly time on a station tion has expired, and the function is not such that the belt is moved intermittently according to the longest time, but the assembly or production belt runs continuously at a suitable average speed.

By introducing bouncing methods in the context of the invention, search space smoothing also represents a further improvement in the solutions already created / described, and also in somewhat different methods, so that this is a fundamental aspect of the invention with also independent inventive step. The same applies to the procedural use and implementation of the bouncing method in principle with existing optimization methods of any kind.

The invention is only illustrated by way of example in the description and in the drawings, and is not limited thereto, but rather includes all variations, modifications, substitutions and combinations which the person skilled in the art can understand the present documents, in particular within the scope of the claims and the general representations in from the introduction to this description and the description of the exemplary embodiments and their representations in the drawing and can combine them with his specialist knowledge and the prior art, in particular with the inclusion of the complete disclosure contents of the older applications specified at the beginning of this description. In particular, all individual features and configuration options of the invention and its exemplary embodiments can be combined.

Claims

Expectations

1. A method for optimizing transportation tasks, wherein at least one location-dependent transportation parameter is recorded, characterized in that probabilities are determined for the at least one location-dependent transportation parameter that are used as a basis for the optimization of the transportation task.

2. Method for optimizing transport tasks according to claim 1, characterized in that, in addition to the probabilities determined for the transport parameters, the optimization of the transport task is used as a basis.

3. A method for optimizing transportation tasks according to claim 1 or 2, characterized in that the probabilities of the at least one location-dependent transportation parameter of the transportation task are used as a basis in the form of an error bar.

4. A method for optimizing transport tasks according to one of the preceding claims, characterized in that the probabilities or possibly error bars are continuously updated or improved taking into account the at least one location-dependent transport parameter.

5. A method for optimizing transport tasks according to one of the preceding claims, characterized in that a temporal variation or dependency of the at least one location-dependent given transportation parameter is recorded, and that time-dependent probabilities are determined for this, which instead of the at least one location and time-dependent given transportation parameter or in addition to the at least one location and time-dependent given transportation parameter of the transportation task.

6. The method for optimizing task tasks according to one of the preceding claims, characterized in that a probability distribution is determined from a plurality of probabilities, which instead of the location-specific and in particular time-dependent given transportation parameter or in addition to the location-based and in particular time-dependent given transportation parameter of the transportation task is placed.

7. The method for optimizing transportation tasks according to one of the preceding claims, characterized in that the location-dependent transportation time is recorded as the at least one location-dependent transportation parameter.

8. A method for optimizing transport tasks according to one of the preceding claims, characterized in that the probabilities determined or, if applicable, the probability distribution determined are used as the basis for route optimization.

9. The method for optimizing transport tasks according to claim 8, characterized in that time-dependent probabilities for the travel times are used, which are constantly adapted to the current traffic situations.

10. Method for optimizing transport tasks according to claim 9, characterized in that the time-dependent probabilities are adapted to the current traffic situation via satellite-based systems.

11. A method for optimizing transport tasks according to claim 9 or 10, characterized in that the adaptation of the time-dependent probabilities to the current traffic situation contains a damping factor which prevents chaotic behavior.

12. A method for optimizing transport tasks according to one of claims 9 to 11, characterized in that satellite-based data from route toll systems are used.

13. A method for optimizing transport tasks according to one of claims 9 to 12, characterized in that inaccurate noisy GPS data are used to create the time-dependent probabilities.

14. A method for optimizing transport tasks according to one of claims 9 to 13, characterized in that for optimization a cost function is expanded by a risk term which reflects the time-dependent probability distributions for the travel times.

15. A method for optimizing transportation tasks according to one of claims 9 to 14, characterized in that for optimizing the probability distributions for the total travel times. Total costs are calculated.

16. The method for optimizing transport tasks according to one of claims 1 to 7, characterized in that the moments of the time-dependent probabilities for cycle times of production processes are used for the optimization.

17. Method for optimizing transport tasks according to one of the preceding claims, characterized in that physical, stochastic, genetic algorithms and / or the savings algorithm are / are used for the optimization.

18. Method for optimizing transport tasks according to one of the preceding claims, characterized in that an analog optimization with probability distributions is used.

19. Method for optimizing transport tasks according to one of the preceding claims, characterized in that probability distributions are used instead of deterministic data.

20. Method for optimizing transport tasks according to one of the preceding claims, characterized in that a bouncing method is used for the optimization.

21. Method for optimizing transport tasks according to one of the preceding claims, characterized in that for optimization simulated annealing, threshold acceptance, great deluge (deluge), genetic algorithms, savings algorithms with or without subsequent bouncing are used.

22. A method for optimizing transport tasks according to claim 20 or 21, characterized in that a bombshell method is used for the optimization.

23. A method for optimizing transport tasks according to claim 22, characterized in that random coordinates are generated in the context of the bombshell method with a random generator, that a radius around each coordinate point is determined or determined or generated, in particular by the random Generator generated and that only those sections of the transport route or the tour that lie within these radii will be optimized with the bouncing method.

24. A method for optimizing transport tasks according to claim 23, characterized in that the random generator repeatedly generates new random coordinates and / or radii around random coordinates.

25. Method for optimizing transport tasks according to one of the preceding claims, characterized in that a risk assessment is included.

26. A method for optimizing transportation tasks according to claim 25, characterized in that the risk assessment is carried out according to the "value-at-risk" method.

27. Method for optimizing transport tasks according to one of the preceding claims, characterized in that search space smoothing methods with or without threshold and / or low-temperature simulated annealing methods are used for optimization.

28. A method for optimizing transport tasks according to one of the preceding claims, characterized in that a small threshold is used when using threshold acceptance or a threshold method.

29. System for optimizing transport tasks, wherein detection devices are provided for detecting at least one location-dependent transportation parameter and information devices that are designed to inform users, including the at least one location-dependent transportation parameter, about an optimization possibility of the transportation task, characterized in that calculation devices are provided, which are designed to determine probabilities for the at least one location-dependent transport parameter, and that the information devices are designed to inform users, including the probabilities, of an option for optimizing the transport task.

30. System for optimizing transport tasks according to claim 29, characterized in that devices for carrying out the method for optimizing transport tasks according to one of claims 2 to 28 are included and can be operated as intended.