Method for supporting operator of ego-agent, program, storage medium, and support system

Abstract

To provide a method and a system for supporting an operator by transmitting a risk in a future state to the operator of an ego-agent.SOLUTION: An action plan algorithm is applied by using a first value and a second value of a parameter in a cost function of the action plan algorithm, for determining first and second planned actions. A present state of the ego-agent is identified, for determining an actual action. A personalized parameter value is estimated on the basis of a relation between the first and second planned actions and the actual action of the ego-agent. A parameter correction value is determined on the basis of the personalized parameter value and a target parameter value. The personalized parameter value is corrected by using the parameter correction value, for generating an adapted parameter value. The action plan algorithm is applied on the basis of the adapted parameter value, and a risk is estimated. The risk is then transmitted to an operator.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present disclosure relates to the field of supporting an operator of an own agent. In particular, a method for supporting an operator of an own agent, a corresponding program including a program code, a corresponding non-transitory computer readable storage medium, and a support system for supporting an operator of an own agent are proposed. [Background technology]

[0002] When people operate their own agents in an environment where other agents are also present, there is generally a risk of collision between the agents. Traffic density has increased significantly in the last years, making it very difficult for the operator of the agent to observe all other agents in order to react in time to the actions of all other agents in the environment so that collisions can be avoided. In addition, there are other risks encountered during the operation of the vehicle, for example, curve risks or regulatory risks. The performance of modern processors and the availability of sensors such as radar sensors, cameras, LIDAR sensors, etc., that provide the processor with information about the environment of the own agent, make it possible to assist the operator of the own agent during the operation of the own agent by the operator. However, the style of operating the own agent varies greatly from operator to operator. Thus, in many situations, the assistance system suggests actions in a way that distracts the operator or even prevents the control actions by the operator (for example, in the case of semi-automated driving).

[0003] There are several approaches that attempt to adapt the system to the individual operator's behavior so that the operator does not feel bothered and therefore the assistance provided by the system is more tolerable. For example, US Patent No. 9,623,878 (B2) proposes a personalized driver assistance system that learns the driver's habits. However, the proposed system adjusts the control parameters, e.g. the target distance of the ACC (adaptive cruise control), to the driver's specific driving style. Unfortunately, such an approach results in an assistance system that reflects the individual driving style but does not influence the driver's behavior.

[0004] A number of different approaches have been developed to generate control signals for autonomous or partially autonomous driving. One approach is described in US Pat. No. 9,463,797 (B2), in which a future trajectory of an own agent is predicted and from the prediction, multiple alternative trajectories of the own agent are generated. Furthermore, a hypothetical future trajectory of another agent is determined and a risk function over time is calculated based on at least one pair of the trajectory of the own agent and the trajectory of the other agent, or alternatives along the calculated hypothetical trajectory of the own agent. These risk functions are then incorporated into a risk map, which is then analyzed to generate a control signal.

[0005] US Patent 10,627,812 (B2) concerns another problem of known assistance systems. To make an accurate prediction, the information about the environment used in the prediction algorithm needs to be accurate. However, since most of the information is obtained using one or more sensors, certain areas in the environment that are relevant for an accurate assessment of the traffic situation may be occluded. US Patent 10,627,812 (B2) reduces the risk of such blind spots by assuming virtual traffic entities in areas where the reliability of the sensor data is below a certain threshold or where no sensor data is available at all, allowing the virtual traffic entities to interact with the predicted behavior of the own agent. A respective risk measure is estimated and the estimated risk measure is taken into account in the control actions of the own agent.

[0006] US Patent Application Publication No. 2020 / 0231149(A1) supports the driver of an ego-agent by considering a priority relationship between the ego-agent and at least one other traffic participant and selecting a prediction model for each of the traffic participants. Thus, the selection of the prediction model takes into account the determined priority relationship between the involved agents, thus improving the assistance by a more accurate prediction of the future actions of the other agents.

[0007] EP 4068153 A1 describes an advanced driver assistance system in which features of the environment are identified based on information received from sensors and risk zones for said features are estimated, and the features and risk zones for said features are then displayed on a display together with the vehicle's environment in a map. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] U.S. Patent No. 9,623,878(B2) [Patent Document 2] U.S. Patent No. 9,463,797 (B2) [Patent Document 3] U.S. Patent No. 10,627,812(B2) [Patent Document 4] US Patent Application Publication No. 2020 / 0231149(A1) Summary of the Invention [Problem to be solved by the invention]

[0009] As the above-provided examples of known assistance systems clearly show, there are several different approaches to estimating the risk and providing information about it to the operator of an agent or to adapt the control of the agent to the habits of that operator. However, there is still the problem that the style of one operator of an agent in a traffic situation may differ significantly from that of another agent. This may lead to a misinterpretation of a particular situation and is difficult not only for human operators but also for automated systems. It is therefore still necessary to find ways to harmonize the operating styles of the operators of several agents involved in a traffic situation. Such a harmonisation significantly increases the quality of the prediction and its reliability, directly improving road safety. [Means for solving the problem]

[0010] In the present invention, this problem is solved by adapting the estimation of future risks based on the habits of the individual operators, but in contrast to what is known from the prior art, by not only trying to meet the operator's expectations but also taking into account the differences between the operator's style and the target style. This is achieved by adapting the estimation of future risks in such a way that the transmitted output is ultimately what educates the operator (driver) towards the target style. This target style is, for example, an average driving style, and if behavior in line with the target style is achieved for every operator of the agents, the dangerous differences in driving styles between different agents can be mitigated.

[0011] According to the present invention, this is achieved by a method, a corresponding program, a computer readable storage medium and an assistance system for assisting an operator by communicating an estimated risk to the operator, the estimation of the risk being adapted to affect the behavior of the own agent starting from an analysis of the driving habits of the operator. First, a behavior planning algorithm is executed, which uses a first value of the parameter and a second value of the parameter in a cost function of the behavior planning algorithm to determine a first and a second planned behavior. Furthermore, a current state of the own agent is identified to derive an actual behavior of the own agent operated by the operator of the own agent. Then, based on a relationship between the first and second planned behaviors and the actual behavior of the own agent, personalized parameter values ​​are estimated, which reflect this actual behavior. By using such personalized parameter values ​​in the behavior planning algorithm, a planned behavior is obtained that corresponds to the habits of the operator of the own agent. Since this does not bring about any change in the habits of the operator, but only models the style of the operator in the cost function used for the behavior planning, the present invention proposes to correct the personalized parameter values ​​using parameter correction values. The determination of the parameter correction values ​​is based on at least the estimated personalized parameter values ​​and the target parameter values. For example, knowing the target parameter values ​​and the personalized parameter values ​​estimated for an individual operator makes it possible to shift the personalized parameter values ​​towards the target parameter values, thereby generating adapted parameter values ​​by correcting the personalized parameter values ​​using the parameter correction values.

[0012] It is also possible to use more than two parameter values ​​to determine more than two planned actions, which can then be used to provide an improved estimation of the personalized parameter values, for example by better interpolation.

[0013] This adapted parameter value is then applied to the action planning algorithm to estimate the risk, which is finally communicated to the operator. Thus, the risk communicated to the operator corresponds more closely to the risk communicated to an operator with an average style. This results in a training effect for the assisted operators, as the risk estimation and communication no longer strictly reflects the operator's style, and ultimately harmonizes the styles of all assisted operators.

[0014] In describing the embodiments, reference is made to the accompanying drawings. [Brief description of the drawings]

[0015] [Figure 1] FIG. 2 illustrates various elements of a cost function used in action planning. [Diagram 2] FIG. 13 is a diagram showing the meaning of a risk map. [Diagram 3] FIG. 13 is a diagram showing the relationship between an action plan based on a risk map and parameter values ​​in a cost function. [Figure 4] FIG. 1 is a block diagram of a process for estimating personalized parameter values. [Diagram 5] FIG. 13 illustrates a process for creating an output that conveys risk based on adapted parameter values. [Figure 6] FIG. 13 is a diagram illustrating the effect of a correction value. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] According to one embodiment, the personalized parameter values ​​are estimated using an interpolation of the actual action with the first and second planned actions. Using interpolation to estimate the personalized parameter values ​​has the advantage that the estimation can be done online, i.e. while the own agent is operating. As a result, the guidance remains valid even if the operator's style changes.

[0017] Furthermore, personalized parameter values ​​can be estimated using a comparison of the acceleration of the actual action with the acceleration of the first and second planned actions. The acceleration is easily perceptible to the ego-agent and requires little or no pre-processing of the sensor values. This reduces computational costs and improves real-time application, for example in vehicles.

[0018] According to one embodiment of the present invention, a parameter correction value is calculated based on the difference between the target parameter value and the personalized parameter value and on a correction factor. This has the advantage that, on the one hand, it directly takes into account how different the actual behavior is from the target behavior, so that a larger correction is made if a larger difference is recognized. On the other hand, the correction value allows for further adjustment as to how strong the correction is. In particular, it is possible to determine the value of the correction factor based on at least one of the amount of difference between the personalized parameter value and the target parameter value and the operator condition. Thus, starting from the difference between the actual behavior and the target behavior and a certain correction factor, many more aspects can be taken into account.

[0019] According to another advantageous embodiment, the estimation of the personalized parameter values ​​is performed repeatedly during the operation of the own agent, which makes it possible to adjust the correction to the actual behavior of the own agent, which may change during the operation of the own agent.

[0020] On the other hand, it may be preferable to estimate the personalized parameter value based on a plurality of actual parameter values ​​estimated based on the relationship between the first and second planned actions and the actual actions of the own agent. By considering a plurality of actual parameter values, it is avoided that the personalized parameter value follows every change in the estimated actual parameter value. For example, a moving average can be calculated from a certain number of actual parameter values, or a hysteresis can be applied to filter small changes in the estimated actual parameter value. In that case, the personalized parameter value is updated only if a significant change in the operator's style is recognized.

[0021] According to another preferred embodiment, the last personalized parameter value estimated during a previous operation of the own agent is used as the personalized parameter value for the risk estimation of the current operation. If the starting point is the personalized parameter value estimated during a previous operation of the own agent, it is particularly preferred to update the last personalized parameter value estimated during a previous operation of the own vehicle with the personalized parameter value estimated during the current operation of the own vehicle and then use the updated personalized parameter value as the personalized parameter value for the risk estimation of the current operation. To enable such use of the previously estimated personalized parameter value, the system comprises a non-volatile memory in which the latest personalized parameter value is stored.

[0022] The actual parameter value may be iteratively estimated based on a relationship between the first and second planned actions and the actual action of the own agent, and a confidence measure may be calculated for the latest actual parameter value. If the confidence measure of this latest actual parameter value exceeds a confidence threshold, this latest actual parameter value is taken over as the personalized parameter value. The confidence threshold may, for example, quantify the frequency of change of the personalized parameter value.

[0023] The parameter value is within an interval, and the lower limit of the interval is used as a first value of the parameter and the upper limit of the interval is used as a second value of the parameter to determine the first and second planned actions. Thus, when the extreme values ​​defined by the limits of the interval are used to determine the first and second planned actions, it is easily possible to use interpolation to estimate personalized parameter values ​​with high accuracy. The interpolation may use a step function, a linear function, two concatenated linear functions, or a sigmoid function to estimate the actual parameter value.

[0024] It should be noted that for embodiments in which no processing of actual parameter values ​​is performed to derive personalized parameters, personalized parameter values ​​and actual parameter values ​​may be used interchangeably. Estimation of personalized parameter values ​​is the same as estimation of actual parameter values. However, in describing embodiments in which actual parameters are processed to determine personalized parameters, a distinction is made between "personalized" and "actual."

[0025] Preferably, the cost function includes at least one of a usefulness factor and a comfort factor in addition to the risk factor, each factor being weighted with a dedicated parameter. Having such a cost function including several different factors makes it possible to guide the operator of the own agent towards a desired behavior not only with respect to the risk aspect but also with respect to other aspects such as usefulness or comfort. In such a case, the method is performed with respect to the risk factor and also with respect to the usefulness factor and / or comfort factor.

[0026] Furthermore, the progression of the correction values ​​over time may be evaluated and feedback regarding the results of the evaluation provided to an operator. The correction values ​​are then stored in a non-volatile memory and an evaluation of the progression of the correction values ​​over time is performed after a predefined time interval or upon request (e.g., from an operator).

[0027] It should be noted that the present invention can be applied to any kind of vehicle, including airplanes, boats, micromobility robots, and even pedestrians. In most cases, the operator of the ego-agent is the driver of the ego-vehicle, the pilot of the airplane, etc. However, the operator may also control the ego-agent from a remote location.

[0028] Before describing all of the estimation and decision steps in more detail, a general background to the invention and the effect of educating the operator will be discussed.

[0029] An operator's individual style in operating his / her agent can be modeled using coefficients that weight individual elements in a cost function used in the action planning. The cost function can be cost=α*risk-β*utility+γ*comfort It can be written as:

[0030] The parameters α, β, and γ adjust the overall cost function and can have values ​​from an interval between a lower bound and an upper bound. Note that the interval does not have to be the same for all parameters. Figure 1 shows the effect of the possible values ​​of the parameters. On the left side, the first element of the cost function is shown, which describes the safety preference.

[0031] On the left side of Fig. 1, a preference for safety is shown. If the first parameter α has a high value, the resulting cost function leads to a planned action that involves braking in advance to reduce the risk. Thus, the first parameter α considers the risk accepted in the planned action. Similarly, the second parameter β defines a preference for utility, which means, for example, that if the value of the second parameter β is set high, the respective planned action allows for a high-speed overtaking. Finally, as shown in the right part of Fig. 1, the third parameter γ is related to a preference for comfort, which, if set to a high value, will result in a planned action that is rather waiting to not change the actual action frequently.

[0032] On the other hand, the cost function and the resulting planned actions when the action planning algorithm is executed using the cost function with the given parameters make it possible to model the behavior style of the operator. Thus, although originally the values ​​of the three parameters define how the action in a certain situation is planned, it is also possible to determine the corresponding parameter values ​​describing the habits and style of the operator from the observed actual behavior of the own agent. For example, the first parameter α describes how close the operator (own agent) approaches the other agent or how long the operator approaches the other agent. The second parameter β describes how fast the operator wants to arrive at the destination or how often the operator changes lanes. Finally, the third parameter γ describes how often the operator accelerates or decelerates or how strong the acceleration or deceleration is.

[0033] Fig. 2 shows the principle of how risk maps are used for action planning. Based on the predicted trajectories 1, 2 of the own agent and another agent, taking into account Gaussian and Poisson distributions for the spatial and temporal uncertainties that may result for example from sensor uncertainties, a risk map 3 can be created in which the speed v is presented over time. The right side of Fig. 2 shows the intersecting trajectories 1, 2 of the own agent 4 and another agent 5, as well as the respective distribution of the positions of the own agent 4 and another agent 5 over time, i.e. along the trajectories 1, 2. These distributions over time of the positions of the agents 4, 5 make it possible to determine the area 6 of a predicted collision between the agents. This predicted collision is contained in the risk map 3 shown on the left side of Fig. 2, which, in addition to the risk, also contains a way of avoiding the risk of the own agent 4. In this illustration, the collision can be avoided by performing a deceleration 7 of the own agent.

[0034] It should be noted that apart from action planning, the risk map is also used to generate warnings and communicate identified risks to the operator of the own agent. Such risk communication or output of warnings and action planning using risk maps is known in the art.

[0035] The trajectory is calculated in a known manner using information about the vehicle states obtained from sensors mounted or carried by the agents, which may be enhanced by inter-agent information. The obtained information is processed by a processor, which performs not only the calculation steps for planning described above, but also the estimation, correction and all other calculations described below in relation to the present invention.

[0036] In the following description, the aspect of "risk" is mainly used for explanation, however, these descriptions are equally valid for estimating the personalized parameter values ​​β, γ of the factors of "usefulness" and / or "comfort".

[0037] The personalized parameter values ​​can be estimated from the cost function given below, based on the risk map used in the action planner as explained above.

[0038] As explained before, the values ​​of the three parameters affect the planned behavior. The first parameter α concerns the risk of collision, especially when another agent is involved. It can be understood as a measure of the size of the risk zone, as shown in the top part of Figure 3. The smaller the value of the first parameter α, the smaller the size of the risk zone and vice versa.

[0039] In the center of Fig. 3, the influence of the value of the second parameter β is shown. This concerns the situation where the own agent 4 is alone, since the value of the second parameter β influences the target velocity. Similarly, the value of the third parameter γ concerns the acceleration. These relationships between the values ​​of the three parameters α, β, γ are exploited to estimate personalized parameter values. In the following description, for the sake of simplicity, only the first parameter α is referred to.

[0040] FIG. 4 illustrates the procedure for estimating the value of the personalized first parameter. First, two risk maps are created based on the actual vehicle state. The first risk map 10 is created using a first value of the first parameter α, and the second risk map 11 is created using a second value of the first parameter α. The parameter values ​​used preferably correspond to the lower limit (e.g., α=0, novice driver) and the upper limit (e.g., α=1, reliable driver) of the interval of values ​​of the first parameter α. Action planning is performed for both created risk maps 10, 11, resulting in two different planned actions 12, 13, labelled in FIG. 4 as "reliable action" and "defensive action".

[0041] An interpolation 14 is then performed based on the two planned actions 12, 13 and the actual actions of the own agent, using the current vehicle state, which is also used to determine the planned actions 12, 13, in order to derive from the vehicle state a current action of the own agent 4 corresponding to the actual control executed by the operator and, as a result, an actual parameter value. It should be noted that if no further processing of the estimates is performed, the actual parameter value is the same as the personalized parameter value. This interpolation may for example be performed by comparing the acceleration values ​​of the planned actions with the acceleration values ​​of the actual actions.

[0042] The bottom right of Figure 4 shows the interval of possible values ​​for the first parameter α and the resulting operator (driver) personalized parameter values ​​for the planned actions obtained from a planning algorithm that uses the lower and upper limits as parameter values.

[0043] Once the personalized parameter values ​​are estimated, a correction is performed to generate adapted parameter values, which are used to estimate risk by generating a risk map using the adaptive parameter values. This process is illustrated in FIG. 5. After the personalized parameter values ​​are determined at 15, parameter correction values ​​are determined and applied to the personalized parameter values ​​at 16. The estimation at 15 may also include processing the actual parameter values ​​to determine the personalized parameter values, as explained above. By correcting the personalized parameter values, adapted parameter values ​​are generated. This adapted parameter value, denoted by α(k) in the figure, is input to a risk model. The risk model generates a warning signal that is communicated to an operator based on the adapted parameter values ​​and the vehicle state. This communication may be a continuous output to the operator, providing a warning on a human-machine interface, or by semi-autonomously adjusting a control input from the operator. The calculation of the risk map and the determination of the risk that is communicated to the operator are no different from the prior art approaches. However, according to the present invention, the parameter values ​​for the application of the risk model or the generation of the risk map are not only adapted to the style of the operator, but are further corrected to achieve an educational effect. For communication, the system is equipped with a display on which the warning or risk measure may be displayed, or alternatively, communication may use a speaker or an audio signal fed to headphones or a similar communication device.

[0044] In addition to the first value of the first parameter and the second value of the first parameter, the personalized parameter value and its shift by the correction value are illustrated in Fig. 6. The figure also illustrates a "normal value" of the first parameter, which is a target value that can be selected to correspond to an average skilled operator. This "normal value" constitutes the target value in the described embodiment. However, this value does not have to be an average across operators. This value can be set to guide all operators towards a safer operation of their agents.

[0045] The adapted parameter values ​​are α adapted (k)=α estimated +(α normal -α estimated )*k It is calculated as:

[0046] k is the personalized parameter value α estimated and the target value α normal The adapted parameter value α is a correction coefficient that allows adjusting how strong the correction is for a given difference between adapted The basis for the calculation of (k) is the estimated parameter value α estimated and the normal parameter value α normal where k can be freely set depending on the desired effect, for example in the interval between 0 and 1. For example, a correction value k=0.1 may slowly lead the driver towards an average safe driving style.

[0047] The correction factor may be adjusted to adjust the strength of the effect. For example, it may be desirable to provide a smaller correction value if the driver is perceived to be less alert, drowsy, and / or stressed. estimated and the normal parameter value α normalSince the difference between is still the same, the correction coefficient k is then adjusted to 0. On the other hand, if the operator is alert and not drowsy or stressed, a stronger correction may be desired. This effect can be achieved by adjusting the coefficient higher (k->1).

[0048] Another way to adapt the correction factor k is to use the estimated parameter value α estimated and the normal parameter value α normal The purpose of the present invention is to take into account the amount of deviation between the normal and the normal style. For example, the correction coefficient k can be increased if the difference is small and reduced if the difference is large. Reducing the correction coefficient k when the difference is large avoids the driver from being overly nervous. On the other hand, increasing the correction coefficient k when the difference is small ensures that the adaptation of the parameter value still has an effect. Otherwise, the closer the operator's style is to the normal style, the more negligible the difference is. A table relating the correction value to the calculated difference can be stored in the memory and retrieved by the processor that performs all the calculations and decisions described above to estimate the risk.

[0049] It should be noted that the description given so far does not distinguish between personalized parameter values ​​and actual parameter values ​​for the own agent's current behavior. This means that the actual parameter values, estimated for example by interpolation as explained above, are directly used as personalized parameter values ​​for further processing and risk estimation communicated to the operator. However, it may be preferable to estimate multiple actual parameter values ​​and determine personalized parameter values ​​on the basis of them. This avoids that variations in the actual parameter values ​​directly affect the risk estimation.

Claims

1. 1. A method for assisting an operator of an agent by communicating risks in future situations to said operator, comprising: applying a behavior planning algorithm using at least a first value of a parameter and a second value of the parameter in a cost function of the behavior planning algorithm to determine at least first and second planned behaviors; Identifying a current state of the agent and determining an actual action to be taken; estimating personalized parameter values ​​based on a relationship between at least the first and second planned actions and the actual actions of the own agent; determining parameter correction values ​​based at least on the personalized parameter values ​​and target parameter values; generating adapted parameter values ​​by correcting the personalized parameter values ​​using the parameter correction values; applying the action planning algorithm based on the adapted parameter values ​​to estimate the future risk; communicating the estimated risk to the operator; A method comprising:

2. The method of claim 1 , wherein the estimation of the personalized parameter value comprises interpolating the actual behavior with at least the first and second planned behaviors.

3. The method of claim 1 , wherein the estimation of the personalized parameter value comprises comparing an acceleration of the actual action with an acceleration of at least the first and second planned actions.

4. The method of claim 1 , wherein the parameter correction values ​​are calculated based on a difference between the target parameter values ​​and the personalized parameter values ​​and a correction factor.

5. 2. The method of claim 1, wherein the parameter correction value is calculated based on a difference between the target parameter value and the personalized parameter value and a correction factor, the value of the correction factor being determined based on at least one of the difference between the personalized parameter value and the target parameter value and an operator condition.

6. The method of claim 1 , wherein the estimation of the personalized parameter values ​​is performed repeatedly during operation of the ego-vehicle.

7. 2. The method of claim 1 , wherein the personalized parameter value is estimated based on a plurality of actual parameter values ​​estimated based on the relationship between at least the first and second planned behaviors and an actual behavior of the own agent.

8. The method of claim 1 , wherein a last personalized parameter value estimated during a previous operation of the ego-vehicle is used as the personalized parameter value for the risk estimation of a current operation.

9. 2. The method of claim 1, wherein a last personalized parameter value estimated during a previous operation of the vehicle is updated with the personalized parameter value estimated during the current operation of the vehicle, and the updated personalized parameter value is used as the personalized parameter value for the risk estimation of the current operation.

10. 2. The method of claim 1 , wherein actual parameter values ​​are iteratively estimated based on the relationship between at least the first and second planned behaviors and the actual behavior of the own agent, a confidence measure is calculated for a latest actual parameter value, and if the confidence measure for the latest actual parameter value exceeds a confidence threshold, the latest actual parameter value takes over as the personalized parameter value.

11. 2. The method of claim 1, wherein the parameter value lies within an interval, a lower limit of the interval being used as the first value of the parameter and an upper limit of the interval being used as the second value of the parameter.

12. The method of claim 1 , wherein the cost function includes, in addition to a risk factor, at least one of a usability factor and a comfort factor, each factor being weighted with a dedicated parameter.

13. 2. The method of claim 1, wherein the cost function includes, in addition to a risk factor, at least one of a usefulness factor and a comfort factor, each factor being weighted with a dedicated parameter, and the method of claim 1 is performed with respect to the risk factor and further with respect to at least one of the usefulness factor and the comfort factor.

14. The method of claim 1 , wherein a progression of the correction value over time is evaluated and feedback is provided to the operator regarding a result of the evaluation.

15. 1. A program comprising program code means for carrying out a method for assisting an operator of an own agent when the program is executed on a computer or digital signal processor, the method comprising: applying a behavior planning algorithm using at least a first value of a parameter and a second value of the parameter in a cost function of the behavior planning algorithm to determine at least first and second planned behaviors; Identifying a current state of the agent and determining an actual action to be taken; estimating personalized parameter values ​​based on a relationship between at least the first and second planned actions and an actual action of the own agent; determining parameter correction values ​​based at least on the personalized parameter values ​​and target parameter values; generating adapted parameter values ​​by correcting the personalized parameter values ​​using the parameter correction values; applying the action planning algorithm based on the adapted parameter values ​​to estimate risk; generating an output signal for communicating the estimated risk to the operator; Including, the program.

16. A non-transitory computer readable storage medium embodying a program of machine readable instructions executable by a digital processing apparatus, the machine readable instructions causing the digital processing apparatus to: applying a behavior planning algorithm using at least a first value of a parameter and a second value of the parameter in a cost function of the behavior planning algorithm to determine at least first and second planned actions; Identifying a current state of the own agent and determining an actual action to be taken; estimating personalized parameter values ​​based on a relationship between at least the first and second planned actions and an actual action of the own agent; and determining parameter correction values ​​based at least on the personalized parameter values ​​and target parameter values; generating adapted parameter values ​​by correcting the personalized parameter values ​​using the parameter correction values; applying the action planning algorithm based on the adapted parameter values ​​to estimate risk; and generating an output signal for communicating the estimated risk to the operator; A non-transitory computer-readable storage medium for causing a program to be executed.

17. A support system for supporting a self agent, comprising: applying a behavior planning algorithm using at least a first value of a parameter and a second value of the parameter in a cost function of the behavior planning algorithm to determine at least first and second planned behaviors; Identifying a current state of the own agent and determining an actual action to be taken; estimating personalized parameter values ​​based on a relationship between at least the first and second planned actions and the actual actions of the own agent; determining parameter correction values ​​based at least on the personalized parameter values ​​and target parameter values; generating adapted parameter values ​​by correcting the personalized parameter values ​​using the parameter correction values; applying the action planning algorithm based on the adapted parameter values ​​to estimate risk; Communicating the estimated risk to the operator The support system includes a processor configured to:

Images