Driving assistance device, vehicle, and driving assistance method

Abstract

A driving assistance device according to one embodiment of the present disclosure, which assists in driving a target vehicle using hypothesis inference, is provided with: an information acquisition unit configured to acquire observation information including one or a plurality of sets of first event information; a hypothesis inference unit configured to infer a plurality of hypotheses, each including one or a plurality of sets of second event information, by applying background knowledge information including a plurality of sets of evidence information to the observation information and thereby performing hypothesis inference; a weight setting unit configured to individually set weights, which indicate degrees of importance, for the sets of evidence information included in the background knowledge information according to the priorities for avoiding the events indicated by the evidence information; and a determination unit configured to determine a maximum likelihood hypothesis, which is the hypothesis having the lowest total cost among the plurality of hypotheses, on the basis of the cost for each set of first event information included in the observation information and the cost for each set of second event information included in each hypothesis.

Description

Driving assistance device, vehicle, and driving assistance method

[0001] The present disclosure relates to a driving assistance device, a vehicle, and a driving assistance method.

[0002] BACKGROUND ART Various techniques have been disclosed as driving assistance devices and driving assistance methods that use abductive reasoning to provide driving assistance for a target vehicle (see, for example, Patent Document 1).

[0003] JP 2016-91039 A

[0004] A driving assistance device according to one embodiment of the present disclosure is a device that provides driving assistance for a target vehicle using abductive reasoning, and includes: an information acquisition unit configured to acquire observation information including one or more pieces of first event information; abductive reasoning unit configured to perform abductive reasoning by applying background knowledge information including a plurality of pieces of evidence information to the observation information, thereby inferring a plurality of hypotheses, each of which includes one or more pieces of second event information; a weight setting unit configured to individually set weights indicating importance for the evidence information included in the background knowledge information in accordance with the avoidance priority of the event indicated by the evidence information; and a determination unit configured to determine a maximum likelihood hypothesis, which is the hypothesis with the lowest total cost, from the plurality of hypotheses based on the cost for each piece of first event information included in the observation information and the cost for each piece of second event information included in each of the hypotheses.

[0005] A vehicle according to an embodiment of the present disclosure is equipped with the driving assistance device according to the embodiment of the present disclosure.

[0006] A driving assistance method according to one embodiment of the present disclosure is a method for providing driving assistance to a target vehicle using abductive reasoning, and includes the steps of: acquiring observation information including one or more pieces of first event information; applying background knowledge information including a plurality of pieces of evidence information to the observation information to perform abductive reasoning, thereby inferring a plurality of hypotheses, each of which includes one or more pieces of second event information; individually setting weights indicating importance for the evidence information included in the background knowledge information in accordance with the avoidance priority of the event indicated by the evidence information; and determining a most likely hypothesis from the plurality of hypotheses, which is the hypothesis with the lowest total cost, based on the cost for each piece of first event information included in the observation information and the cost for each piece of second event information included in each of the hypotheses.

[0007] The accompanying drawings are included to provide a further understanding of the disclosure, and are incorporated in and constitute a part of this specification. The drawings illustrate one embodiment and, together with the description, serve to explain the principles of the disclosure.

[0008] FIG. 1 is a schematic diagram illustrating an example of a general configuration of a vehicle equipped with a driving assistance device according to an embodiment of the present disclosure. FIG. 2 is a block diagram illustrating an example of a detailed configuration of the vehicle illustrated in FIG. 1. FIG. 3 is a flowchart illustrating an example of a driving assistance process according to an embodiment. FIG. 4 is a schematic diagram illustrating an overview of the abductive reasoning illustrated in FIG. 3. FIG. 5 is another schematic diagram illustrating an overview of the abductive reasoning illustrated in FIG. 3. FIG. 6 is a diagram illustrating a specific example of a weight setting process illustrated in FIG. 3. FIG. 7A is a diagram illustrating an example of a correspondence relationship between the occurrence frequency and weight of a immediately preceding event illustrated in FIG. 6. FIG. 7B is a diagram illustrating an example of a correspondence relationship between the occurrence margin time of a immediately preceding event and weight illustrated in FIG. 6. FIG. 8 is a schematic diagram illustrating an example of setting weights for other events illustrated in FIG. 6. FIG. 9 is a schematic diagram illustrating an example of a process for abductive reasoning according to Comparative Example 1 and Example 1. FIG. 10 is a schematic diagram illustrating an example of a process for abductive reasoning according to Comparative Example 2 and Example 2.

[0009] In a driving assistance device or the like that provides driving assistance for a target vehicle using abductive reasoning, for example, it is required to improve the accuracy of determining the maximum likelihood hypothesis. It is desirable to provide a driving assistance device, a vehicle, and a driving assistance method that can improve the accuracy of determining the maximum likelihood hypothesis.

[0010] Some exemplary embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. Note that the following description illustrates one specific example of the present disclosure and should not be construed as limiting the present disclosure. For example, each element, including numerical values, shapes, materials, parts, the position of each part, and the connection method of each part, is merely an example and should not be construed as limiting the present disclosure. Furthermore, in the following exemplary embodiments, components not described in independent claims based on the highest concept of the present disclosure are optional and may be provided as needed. The drawings are schematic and are not intended to be drawn to scale. Throughout this specification and the drawings, components having substantially the same function and configuration are designated by the same reference numerals, and redundant description will be omitted. Furthermore, components not directly related to one embodiment of the present disclosure are not shown in the drawings.

[0011] 1. Embodiment> [Configuration] Fig. 1 is a schematic diagram illustrating an example of the overall configuration of a vehicle 1 equipped with a driving assistance device (a driving assistance device 111 described later) according to an embodiment of the present disclosure. Fig. 2 is a block diagram illustrating an example of the detailed configuration of the vehicle 1 illustrated in Fig. 1.

[0012] 1 and 2, the vehicle 1 includes a vehicle control unit 11, a battery 12, a communication device 13, and a camera 14. As shown in FIG. 1, the vehicle 1 has an occupant 9 (such as a driver) on board.

[0013] (A. Vehicle Control Unit 11) The vehicle control unit 11 is a component (control unit) that controls various operations in the vehicle 1 and performs various arithmetic processing. Specifically, the vehicle control unit 11 includes, for example, one or more processors (CPU: Central Processing Unit) that execute programs, and one or more memories communicatively connected to these processors. Furthermore, such memories include, for example, RAM (Random Access Memory) that temporarily stores processing data, and ROM (Read Only Memory) that stores programs.

[0014] In the example shown in FIG. 2, the vehicle control unit 11 includes a driving assistance device 111, a battery control unit 112, a communication control unit 113, and a camera control unit 114.

[0015] (A-1. Driving assistance device 111) The driving assistance device 111 is a device that provides driving assistance to a target vehicle (vehicle 1) using hypothetical reasoning, which will be described later, and in the example shown in FIG. 2, includes an information acquisition unit 21, a hypothetical reasoning unit 22, a weight setting unit 23, a determination unit 24, a notification unit 25, and a driving control unit 26.

[0016] The notification unit 25 and the driving control unit 26 each correspond to a specific example of a "control unit" in an embodiment of the present disclosure.

[0017] The information acquisition unit 21 is a unit that acquires various types of information related to the vehicle 1. Specifically, in the present embodiment, the information acquisition unit 21 acquires, as one type of such various information, observation information Im including one or more pieces of first event information Ie1, which will be described later, using the camera 14, various sensors, etc. Details of such observation information Im will be described later.

[0018] The hypothetical reasoning unit 22 is a unit that performs hypothetical reasoning by applying background knowledge information Ibk, which includes a plurality of pieces of evidence information Ir (described later), to the above-mentioned observation information Im, to infer a plurality of hypotheses H, each of which includes one or a plurality of pieces of second event information Ie2 (described later). Details of such hypothetical reasoning processing will be described later.

[0019] The weight setting unit 23 is a unit that individually sets a weight Wr indicating importance for each piece of evidence information Ir included in the background knowledge information Ibk, in accordance with the avoidance priority Pa of the event indicated by the piece of evidence information Ir. Details of the avoidance priority Pa and the process of setting the weight Wr will be described later.

[0020] The determination unit 24 is a unit that determines a most likely hypothesis Hm, which is the hypothesis H with the lowest total cost among the multiple hypotheses H, based on the cost C for each piece of first event information Ie1 included in the above-mentioned observation information Im and the cost C for each piece of second event information Ie2 included in each of the above-mentioned hypotheses H. Details of the process of determining such a most likely hypothesis Hm will be described later.

[0021] The notification unit 25 is a unit that performs a predetermined notification operation (notification processing) for the occupant 9 (such as the driver) of the vehicle 1. The notification unit 25 is configured to include, for example, a display unit (such as a display of various types) that performs a predetermined display operation as such a notification operation, and a tone output unit (such as a speaker of various types) that performs a predetermined audio output operation as such a notification operation.

[0022] The driving control unit 26 is a unit that controls the driving operation of the vehicle 1 (performs driving control processing), and performs overall control regarding the driving of the vehicle 1. Specifically, the driving control unit 26 controls, for example, the drive system, braking system, steering system, and the like of the vehicle 1.

[0023] Here, the notification unit 25 and the driving control unit 26 of the present embodiment each use the maximum likelihood hypothesis Hm determined by the determination unit 24 (based on the second event information Ie2 included in the maximum likelihood hypothesis Hm) to provide driving assistance for the vehicle 1. Specifically, the notification unit 25 performs notification processing using such maximum likelihood hypothesis Hm, and the driving control unit 26 performs driving control processing using such maximum likelihood hypothesis Hm.

[0024] (A-2. Battery control unit 112, communication control unit 113, camera control unit 114) The battery control unit 112 is a unit that controls the operation (charging operation, discharging operation, etc.) of the battery 12. The communication control unit 113 is a unit that controls the communication operation of the communicator 13. The camera control unit 114 is a unit that controls the imaging operation of the camera 14.

[0025] (B. Battery 12, Communicator 13, Camera 14) The battery 12 is a component that functions as a power source for the vehicle 1, and is configured using various types of secondary batteries such as lithium ion batteries. The communicator 13 is a device that performs various types of communication with the outside of the vehicle 1 (for example, an information processing device such as a server provided outside the vehicle 1, or other vehicles other than the vehicle 1).

[0026] The camera 14 is a component that acquires imaging data relating to the outside (such as the surrounding environment of the vehicle 1) and the inside (such as the occupants of the vehicle 1) of the vehicle 1. The imaging data acquired in this manner is supplied to, for example, the driving assistance device 111 (such as the information acquisition unit 21 described above) in the vehicle control unit 11.

[0027] [Operation, Function, and Effect] Next, the operation, function, and effect of this embodiment will be described in detail.

[0028] Fig. 3 is a flowchart showing an example of driving assistance processing according to this embodiment (a processing example in the driving assistance device 111). Figs. 4 and 5 are each a schematic diagram for explaining an outline of the hypothetical reasoning (step S3, described later) shown in Fig. 3. Note that Fig. 5 shows a situation in which, on a road having road signs St (green signs) on each of three driving lanes L1 to L3, a preceding vehicle 7 and a stopped vehicle 8 (stopped ahead of the preceding vehicle 7) are present ahead of vehicle 1 (traveling in driving lane L2), which is the target vehicle of the driving assistance processing.

[0029] (A. Overview of Hypothetical Reasoning) Before describing the overall driving support process shown in FIG. 3, an overview of hypothetical reasoning will be described.

[0030] First, abductive reasoning is a method of inferring premises by applying rules to a conclusion. On the other hand, deduction is a method of inferring a conclusion from rules and premises. As such, abductive reasoning and deduction are characterized by different directions of reasoning, with abductive reasoning also being called backward reasoning and deduction also being called forward reasoning. When such abductive reasoning is applied to vehicle driving assistance processing (hazard prediction), it is as follows. That is, the above-mentioned hypothesis H (corresponding to a "premises") is inferred by applying the above-mentioned previously set background knowledge information Ibk (corresponding to a "rule") to the above-mentioned observation information Im and events indicating a hazard (corresponding to a "conclusion").

[0031] Furthermore, the degree to which the hypothesis H inferred in this way matches the observation information Im (goodness of proof) can be output as a cost C. In other words, the lower this cost C is, the better the proof is. When outputting this cost C, the following two types of "weights" are used:

[0032] The "weight" indicating the importance of each piece of evidence information Ir is registered in the background knowledge information Ibk, and the more important the piece of evidence information Ir is, the higher the "weight" is set. The "weight" indicating the reliability of each piece of information (first event information Ie1) included in the observation information Im is registered in the observation information Im, and the more reliable the first event information Ie1 is, the lower the "weight" is set.

[0033] Here, a specific example of such a hypothesis overview processing will be described with reference to Figures 4 and 5. Note that the content of each piece of information and the value of cost C shown in Figure 4 are merely examples, and any settings are possible. Also, the symbol "^" shown in Figure 4 indicates a logical product ("AND"). These points also apply to Figures 7A to 10, which will be described later.

[0034] First, in the observation information Im obtained by various sensors, etc., a cost C is assigned to each piece of first event information Ie1 according to the weight indicating the reliability described above. In the example shown in FIG. 4, a cost C of $10 is assigned to each piece of first event information Ie1 indicating "vehicle ahead" and "green sign ahead." In addition, in the example shown in FIG. 4, the first event information Ie1 indicating "occurrence of danger" is not directly observed and is therefore uncertain (low reliability), so its cost C is set to a value 10 times the above (cost C = $100). In the example shown in FIG. 5, this "occurrence of danger" refers to an event in which vehicle 1 comes into contact with a preceding vehicle 7 (see arrow P1 in FIG. 5).

[0035] Next, a hypothesis H is inferred by applying background knowledge information Ibk to such observation information Im and performing abductive reasoning. In the example shown in FIG. 4 , the background knowledge information Ibk includes a plurality of pieces of evidence information Ir (evidence information Ir indicating "there is a preceding vehicle," "there is a stopped vehicle," and "there is a green sign"), and the weights indicating the above-mentioned importance (weights Wr, which will be described later) are set to 0.3, 0.6, and 0.3, respectively. Also, in the example shown in FIG. 4 , hypothesis H includes a plurality of pieces of second event information Ie2 (second event information Ie2 indicating "there is a preceding vehicle," "there is a stopped vehicle," and "there is a green sign").

[0036] Here, in the example shown in FIG. 4, the total cost (total value of costs C) for hypothesis H obtained in this way is $120 (= $30 + $60 + $30), which is 1.2 times the value before inference (cost C for "risk occurrence" = $100). In other words, because hypothesis H is not observational information, the total cost increases. The total cost thus increased is allocated to each of the multiple pieces of second event information Ie2, which are factors (literals) that make up hypothesis H. At this time, the importance (weight Wr) of each piece of evidence information Ir described above is reflected.

[0037] Next, if the same type of information (literal) exists between the first event information Ie1 included in the observation information Im and the second event information Ie2 included in the hypothesis H, the following unification process is performed. That is, in this unification process, between the first event information Ie1 and the second event information Ie2, the information with the relatively larger cost C is deleted, and only the information with the relatively smaller cost C is adopted. In the example shown in FIG. 4 , the unification process indicated by symbol U1 adopts the information with the relatively smaller cost C (first event information Ie1 indicating "there is a preceding vehicle"). Furthermore, the unification process indicated by symbol U2 adopts the information with the relatively smaller cost C (first event information Ie1 indicating "there is a green traffic sign").

[0038] In this way, the most likely hypothesis Hm is determined, which is the hypothesis H with the lowest total cost (the best evidence). In the example shown in Figure 4, the most likely hypothesis Hm includes multiple pieces of second event information Ie2 (second event information Ie2 indicating "vehicle ahead," "green sign," and "parked vehicle"), and has a total cost of $80 (= $10 + $10 + $60).

[0039] (B. Driving Assistance Processing) Next, based on the outline of the abductive reasoning described above, the driving assistance processing of this embodiment shown in FIG. 3 will be described in detail.

[0040] 3, first, the information acquisition unit 21 acquires (step S1) observation information Im including the first event information Ie1 as one of various pieces of information related to the vehicle 1. Note that the observation information Im may include not only current information but also, for example, past information (such as information from the past few seconds).

[0041] (S2: Setting Process of Weight Wr) Next, the weight setting unit 23 individually sets a weight Wr indicating importance for each piece of evidence information Ir included in the background knowledge information Ibk, in accordance with the avoidance priority Pa of the event indicated by that piece of evidence information Ir (step S2). As shown in FIG. 6, the weight setting unit 23 may update the weight setting value for each piece of evidence information Ir, for example, periodically (daily, for example). In the example series of processes shown in FIG. 3, the setting process of weight Wr is performed after the process of acquiring observation information Im (step S1). However, the setting process of weight Wr may also be performed in advance, for example, before the process of acquiring observation information Im.

[0042] Here, Fig. 6 is a diagram for explaining a specific example of the process for setting such a weight Wr. Fig. 7A shows an example of the correspondence relationship between the occurrence frequency fx of the immediately preceding event E1 (described later) shown in Fig. 6 and the weight Wx (Wr1). Fig. 7B shows an example of the correspondence relationship between the occurrence margin time tx of the immediately preceding event E1 (described later) and the weight Wx (Wr1). Fig. 8 is a schematic diagram for explaining an example of setting the weight Wr2 for another event E2 (described later) shown in Fig. 6.

[0043] First, in this embodiment, the weight setting unit 23 performs the setting process of the weight Wr described above using the following three methods (methods (A) to (C)) (see FIG. 6).

[0044] Method (A): The weight Wr is set according to the level Pa1 of the urgency Em of the event as the avoidance priority Pa described above. Method (B): The weight Wr is set according to the level Pa2 of the encounter frequency fe of the event as the avoidance priority Pa described above. Method (C): The weight Wr is set according to both the level Pa1 of the urgency Em and the level Pa2 of the encounter frequency fe described above.

[0045] Specifically, in the above-described method (A), the weight setting unit 23 sets a relatively high weight Wr=Wr1 corresponding to the immediately preceding event E1, which is an event with a relatively high urgency Em. In the above-described method (B), the weight setting unit 23 sets a relatively high weight Wr corresponding to an event with a relatively high encounter frequency fe (high-frequency event). In the above-described method (C), the weight setting unit 23 prioritizes the setting of the weight Wr according to the level Pa1 of urgency Em over the setting of the weight Wr according to the level Pa2 of encounter frequency fe.

[0046] The preceding event E1 refers to an event that occurs immediately before the occurrence of a hazard (for example, several seconds or more before the occurrence of the hazard). Examples of such preceding events E1 include braking, turning on a blinker, and cutting in on the side of the vehicle.

[0047] 7A and 7B, the weight setting unit 23 sets the weight Wr1 corresponding to the immediately preceding event E1 in accordance with at least one of the occurrence frequency fx and the time to occurrence tx of the immediately preceding event E1. Note that the time to occurrence tx means, for example, the average value of a value (time) indicating how many seconds before the occurrence of the hazard the immediately preceding event E1 occurred.

[0048] 7A and 7B, the weight Wx set according to the occurrence frequency fx and the weight Wx set according to the occurrence margin time tx are defined as shown in the following equations (1) and (2). Note that Wmax, Wmin, and tmax in these equations (1) and (2) respectively indicate the maximum and minimum values ​​of the weight Wx and the maximum value of the occurrence margin time tx.

[0049]

[0050]

[0051] Specifically, in the example shown in Figure 7A, the three immediate preceding events E1 are "braking," "turn signal," and "pulling closer to the side," and the occurrence frequencies fx = fb, ft, and fs of these three immediate preceding events E1 are fb:ft:fs = 2:1:1, respectively. Therefore, the weight Wx of each immediate preceding event E1 is set according to the occurrence frequency fx of each immediate preceding event E1 (the weight Wx is set so that it increases as the occurrence frequency fx increases). That is, in the example of Figure 7A, the weights Wx = Wb, Wt, and Ws of "braking," "turn signal," and "pulling closer to the side" are set to Wb = 0.8, Wt = 0.7, and Ws = 0.7, respectively.

[0052] In the example shown in Figure 7B, the three most recent events E1 are "braking," "turn signal," and "pulling over," and the occurrence margin times tx = tb, tc, and ts for these three most recent events E1 are tb = 2.0 [s], tt = 1.0 [s], and ts = 0.5 [s], respectively. Therefore, the weight Wx of each most recent event E1 is set according to the occurrence margin time tx of each of these most recent events E1 (the weight Wx is set so that it increases as the occurrence margin time tx becomes shorter). That is, in the example of Figure 7B, the weights Wx = Wb, Wt, and Ws for "braking," "turn signal," and "pulling over," are set to Wb = 0.6, Wt = 0.8, and Ws = 0.9, respectively.

[0053] The occurrence frequency fx and occurrence margin time tx of such a previous event E1 are each obtained (measured) by the information acquisition unit 21 using the camera 14 and various sensors, for example, as follows.

[0054] - Measurement data on the road structure and traffic participants around vehicle 1 can be acquired using stereo cameras, LiDAR (Light Detection And Ranging), side radar, etc., and image recognition processing and object detection processing can be performed based on the acquired measurement data, making it possible to recognize the surrounding road structure and traffic participants. - The position of vehicle 1, the driving lane, the presence or absence of intersections, etc. can be recognized based on GPS (Global Positioning System), high-precision map information, and image matching. - By using vehicle-to-vehicle communication and road-to-vehicle communication based on surrounding traffic environment information acquired using sensors such as surveillance cameras installed on the road and external camera systems on vehicles other than vehicle 1, it is possible to recognize information that cannot be obtained from vehicle 1's sensors or map information (for example, information that is blocked by other vehicles, such as the number of other vehicles and other vehicles on the vehicle's route). - Various states of vehicle 1 can be acquired using vehicle 1's CAN (Controller Area Network) and IMU (Inertial Measurement Unit).

[0055] 6, the weight setting unit 23 sets the weight Wr1 corresponding to the immediately preceding event E1 to a value equal to or greater than half of the total value Wr(total) of the weights Wr in the background knowledge information Ibk (Wr1≧Wr(total)×0.5). Furthermore, the weight setting unit 23 distributes and sets the weight Wr2 (=the total value Wr(total)−weight Wr1 of the immediately preceding event E1) for other events E2 other than the immediately preceding event E1 according to the encounter frequency fe for each of the other events E2 (see FIG. 6).

[0056] 8, the weight Wr (Wr1=0.6) corresponding to "braking" as the immediately preceding event E1 is set to half the total weight Wr (Wr(total)) (=0.4+0.2+0.6=1.2) in the background knowledge information Ibk (Wr1=Wr(total)×0.5). Also, in the example of FIG. 8, the weights Wr2 corresponding to events E2 other than the immediately preceding event E1 (weights Wr=0.4, 0.2 corresponding to "branch" and "mirror blind spot") are distributed according to the encounter frequency fe (=100%, 50%) of each of the other events E2, in other words, the encounter ratio Re (=2:1).

[0057] (S3: Hypothetical Reasoning Process) Subsequently, the hypothetical reasoning unit 22 performs the above-described hypothetical reasoning to infer multiple hypotheses H (step S3). Specifically, the hypothetical reasoning unit 22 performs hypothetical reasoning by applying the background knowledge information Ibk to the observation information Im acquired in step S1, thereby inferring multiple hypotheses H including the second event information Ie2.

[0058] (S4: Process for Determining Maximum Likelihood Hypothesis Hm) Next, the determiner 24 performs the process for determining the maximum likelihood hypothesis Hm described above (step S4). Specifically, the determiner 24 determines the maximum likelihood hypothesis Hm, which is the hypothesis H with the lowest total cost among the multiple hypotheses H, based on the cost C for each piece of first event information Ie1 included in the observation information Im and the cost C for each piece of second event information Ie2 included in the multiple hypotheses H obtained in step S3.

[0059] (S5: Driving Assistance Processing) Next, the notification unit 25 and the driving control unit 26 each use the maximum likelihood hypothesis Hm determined in step S4 to provide driving assistance for the vehicle 1 (step S5). Specifically, the notification unit 25 uses such maximum likelihood hypothesis Hm to perform the various notification processing described above, and the driving control unit 26 uses such maximum likelihood hypothesis Hm to perform the various driving control processing described above.

[0060] This completes the series of processing steps shown in FIG.

[0061] (C. Examples) Next, specific examples (Examples 1 and 2) of the driving assistance process (the process during the above-described hypothetical reasoning) of this embodiment will be described in detail while comparing them with comparative examples (Comparative Examples 1 and 2).

[0062] Fig. 9 is a schematic representation of an example of processing during abductive reasoning according to Comparative Example 1 and Example 1. Fig. 10 is a schematic representation of an example of processing during abductive reasoning according to Comparative Example 2 and Example 2. Note that, for the sake of convenience, some of the symbols for various information and cost C are omitted in Figs. 9 and 10.

[0063] (C-1. Comparative Example 1, Example 1) In both Comparative Example 1 and Example 1 shown in FIG. 9 , the observation information Im includes a plurality of pieces of first event information Ie1 (first event information Ie1 indicating a "junction," a "preceding vehicle," and "braking"). The cost C for each of these pieces of first event information Ie1 is set to $10. In addition, two types of background knowledge information Ibk applied to "occurrence of a risk" are provided: background knowledge information Ibk1 (information corresponding to a "lane change") and background knowledge information Ibk2 (information corresponding to a "rear-end collision"). The background knowledge information Ibk1 includes a plurality of pieces of evidence information Ir (evidence information Ir indicating a "junction," a "blind spot in the mirror," and "braking"), and the background knowledge information Ibk2 includes a plurality of pieces of evidence information Ir (evidence information Ir indicating a "junction," a "preceding vehicle," and "braking of the preceding vehicle").

[0064] (Comparative Example 1) Here, in Comparative Example 1 shown in Figure 9, the weights Wr of each piece of evidence information Ir contained in background knowledge information Ibk1 and Ibk2 are set to be equal to each other (cost C of each piece of evidence information Ir = $40).

[0065] In this comparative example 1, when the unification processes indicated by the reference symbols U11 and U12 are performed on a hypothesis H1 obtained by abductive reasoning using background knowledge information Ibk1, the information indicating the "branch point," "preceding vehicle," "brake," and "mirror blind spot" is included as second event information Ie2. Therefore, the total cost C(H1) for this hypothesis H1 is $70 (= $10 + $10 + $10 + $40). Furthermore, when the unification processes indicated by the reference symbols U21 and U22 are performed on a hypothesis H2 obtained by abductive reasoning using background knowledge information Ibk2, the information indicating the "branch point," "preceding vehicle," "brake," and "front-preceding vehicle brake" is included as second event information Ie2. Therefore, the total cost C(H2) in this hypothesis H2 is also $70 (= $10 + $10 + $10 + $40).

[0066] In this way, in Comparative Example 1, the total costs of hypothesis H1 obtained using background knowledge information Ibk1 (information corresponding to "lane change") and hypothesis H2 obtained using background knowledge information Ibk2 (information corresponding to "rear-end collision") are equal (C(H1) = C(H2)).

[0067] 9 , unlike the above-described comparative example 1, the weights Wr of the pieces of evidence information Ir included in the background knowledge information Ibk1 and Ibk2 are individually set according to the avoidance priority Pa. Specifically, for the pieces of evidence information Ir indicating "junction," "mirror blind spot," and "braking" included in the background knowledge information Ibk1, costs C are set to $40, $20, and $60, respectively, based on the individually set weights Wr. Furthermore, for the pieces of evidence information Ir indicating "junction," "preceding vehicle," and "front-preceding vehicle braking" included in the background knowledge information Ibk2, costs C are set to $20, $40, and $60, respectively, based on the individually set weights Wr as described above.

[0068] In Example 1, when the unification processes indicated by the reference symbols U11 and U12 are performed on hypothesis H1 obtained by abductive reasoning applying background knowledge information Ibk1, the information indicating the "branch point," "preceding vehicle," "braking," and "mirror blind spot" is included as second event information Ie2. However, unlike Comparative Example 1 described above, in Example 1, the total cost C(H1) for hypothesis H1 is $50 (= $10 + $10 + $10 + $20). Similarly, when the unification processes indicated by the reference symbols U21 and U22 are performed on hypothesis H2 obtained by abductive reasoning applying background knowledge information Ibk2, the information indicating the "branch point," "preceding vehicle," "braking," and "front-preceding vehicle braking" is included as second event information Ie2. However, in this case, unlike the case of Comparative Example 1 described above, the total cost C(H2) in this hypothesis H2 is $90 (=$10+$10+$10+$60).

[0069] Thus, in Example 1, the total costs of hypothesis H1 obtained using background knowledge information Ibk1 (information corresponding to "lane change") and hypothesis H2 obtained using background knowledge information Ibk2 (information corresponding to "rear-end collision") are different from each other. Specifically, in the example shown in FIG. 9 , C(H1)<C(H2), and the total cost of hypothesis H1 obtained based on background knowledge information Ibk1 is lower than that of hypothesis H2 obtained based on background knowledge information Ibk2. Therefore, in this case, hypothesis H1 is determined as the most likely hypothesis Hm. In other words, in Example 1, unlike Comparative Example 1, the hypothesis H1 corresponding to a risk with a relatively high urgency Em ("lane change" in the example of FIG. 9 ) can be output as the most likely hypothesis Hm (the hypothesis H with the lowest total cost).

[0070] (C-2. Comparative Example 2, Example 2) Next, in both Comparative Example 2 and Example 2 shown in FIG. 10, the observation information Im includes a plurality of pieces of first event information Ie1 (first event information Ie1 indicating "mirror blind spot" and "preceding vehicle"). The cost C for each of these pieces of first event information Ie1 is set to $10. Furthermore, as in the example of FIG. 9, two types of background knowledge information Ibk that are applied to "occurrence of danger" are provided: background knowledge information Ibk1 (information corresponding to "lane change") and background knowledge information Ibk2 (information corresponding to "rear-end collision").

[0071] (Comparative Example 2) Here, in Comparative Example 2 shown in Figure 10, as in Comparative Example 1 described above, the weights Wr for each piece of evidence information Ir contained in the background knowledge information Ibk1 and Ibk2 are set to be equal to each other (cost C of each piece of evidence information Ir = $40).

[0072] In Comparative Example 2, when the unification processes indicated by the symbol U1 are performed on hypothesis H1 obtained by abductive reasoning using background knowledge information Ibk1, the information indicating the "mirror blind spot," "preceding vehicle," "junction," and "braking" is included as second event information Ie2. The total cost C(H1) for this hypothesis H1 is $100 (= $10 + $10 + $40 + $40). Similarly, when the unification processes indicated by the symbol U2 are performed on hypothesis H2 obtained by abductive reasoning using background knowledge information Ibk2, the information indicating the "mirror blind spot," "preceding vehicle," "junction," and "braking of the preceding vehicle" is included as second event information Ie2. The total cost C(H2) for this hypothesis H2 is also $100 (= $10 + $10 + $40 + $40).

[0073] In this way, in comparison example 2, the total costs of hypothesis H1 obtained using background knowledge information Ibk1 (information corresponding to "lane change") and hypothesis H2 obtained using background knowledge information Ibk2 (information corresponding to "rear-end collision") are equal (C(H1) = C(H2)).

[0074] 10 , unlike the comparative example 2 described above, the weights Wr of the pieces of evidence information Ir included in the background knowledge information Ibk1 and Ibk2 are individually set according to the avoidance priority Pa. Specifically, the costs C of the pieces of evidence information Ir indicating "junction," "mirror blind spot," and "braking" included in the background knowledge information Ibk1 are set to $35, $25, and $60, respectively, based on the individually set weights Wr. Furthermore, the costs C of the pieces of evidence information Ir indicating "junction," "preceding vehicle," and "front-preceding vehicle braking" included in the background knowledge information Ibk2 are set to $20, $40, and $60, respectively, based on the individually set weights Wr as described above.

[0075] In Example 2, when the unification processes indicated by the symbol U1 are performed on hypothesis H1 obtained by abductive reasoning using background knowledge information Ibk1, the information indicating the "mirror blind spot," "preceding vehicle," "junction," and "braking" is included as second event information Ie2. However, unlike Comparative Example 2 described above, in Example 2, the total cost C(H1) for hypothesis H1 is $115 (= $10 + $10 + $35 + $60). Similarly, when the unification processes indicated by the symbol U2 are performed on hypothesis H2 obtained by abductive reasoning using background knowledge information Ibk2, the information indicating the "mirror blind spot," "preceding vehicle," "junction," and "braking of the preceding vehicle" is included as second event information Ie2. In this case, the total cost C(H2) for hypothesis H2 is $100 (=$10+$10+$20+$60).

[0076] Thus, in Example 2, the total costs of hypothesis H1 obtained using background knowledge information Ibk1 (information corresponding to "lane change") and hypothesis H2 obtained using background knowledge information Ibk2 (information corresponding to "rear-end collision") are different from each other. Specifically, in the example shown in FIG. 10 , C(H1)>C(H2), and the total cost of hypothesis H2 obtained based on background knowledge information Ibk2 is lower than that of hypothesis H1 obtained based on background knowledge information Ibk1. Therefore, in this case, hypothesis H2 is determined as the most likely hypothesis Hm. In other words, in Example 2, unlike Comparative Example 2, when no event with a high urgency Em is observed, hypothesis H2 corresponding to a risk with a relatively high encounter frequency fe (in the example of FIG. 10 , "rear-end collision") can be output as the most likely hypothesis Hm.

[0077] (D. Functions and Effects) In this embodiment, multiple hypotheses H are inferred by applying background knowledge information Ibk to observation information Im to perform hypothetical inference. Furthermore, a most likely hypothesis Hm from among the multiple hypotheses H is determined based on the cost C for each piece of first event information Ie1 included in the observation information Im and the cost C for each piece of second event information Ie2 included in each hypothesis H. When determining the most likely hypothesis Hm in this manner, a weight Wr indicating importance is reflected in each piece of evidence information Ir included in the background knowledge information Ibk, the weight Wr being individually set in accordance with the avoidance priority Pa of the event indicated by each piece of evidence information Ir. This makes it possible to determine an appropriate hypothesis H in accordance with the avoidance priority Pa of each event as the most likely hypothesis Hm. As a result, this embodiment makes it possible to improve the accuracy with which the most likely hypothesis Hm is determined.

[0078] Furthermore, in this embodiment, the weight Wr is set according to the level Pa1 of the urgency Em of the event and the level Pa2 of the encounter frequency fe as the avoidance priority Pa, as follows: That is, by setting the weight Wr1 corresponding to the immediately preceding event E1, which is an event with a relatively high urgency Em, relatively high, or by setting the weight Wr corresponding to an event with a relatively high encounter frequency fe, relatively high, it becomes possible to more appropriately determine the maximum likelihood hypothesis Hm. As a result, it becomes possible to further improve the accuracy of determining the maximum likelihood hypothesis Hm.

[0079] Furthermore, in this embodiment, if the weight Wr corresponding to the level Pa1 of the urgency Em is given priority over the weight Wr corresponding to the level Pa2 of the encounter frequency fe, the maximum likelihood hypothesis Hm can be determined more appropriately, thereby further improving the accuracy of determining the maximum likelihood hypothesis Hm.

[0080] Additionally, in this embodiment, when the weight Wr1 corresponding to the immediately preceding event E1 is set in accordance with at least one of the occurrence frequency fx and the time to occurrence tx of the immediately preceding event E1, the following occurs: In other words, the weight Wr1 corresponding to the immediately preceding event E1 can be set more appropriately, which makes it possible to further improve the accuracy of determining the maximum likelihood hypothesis Hm.

[0081] 2. Modifications The present disclosure has been described above by giving several embodiments and examples, but the present disclosure is not limited to these embodiments and can be modified in various ways.

[0082] For example, the configuration (type, arrangement, number, etc.) of each component in a vehicle or the like is not limited to that described in the above embodiment, etc. In other words, the configuration of each component may be of a different type, arrangement, number, etc. Specifically, for example, in the above embodiment, etc., an example has been described in which the driving assistance device in one embodiment of the present disclosure is provided in a target vehicle for driving assistance, but this example is not limiting. In other words, at least some of the components in the driving assistance device (such as the information acquisition unit, the abductive reasoning unit, the weight setting unit, and the determination unit) may be provided, for example, inside a server (information processing device) or the like provided outside the target vehicle.

[0083] Furthermore, in the above embodiments, various processing examples (such as the overall driving assistance processing described above, the processing for setting the weight Wr, the abductive reasoning processing, and the processing for determining the maximum likelihood hypothesis Hm) have been specifically described, but the methods are not limited to those described in the above embodiments, and other methods may be used, for example.

[0084] Furthermore, the values, ranges, magnitude relationships, etc. of the various parameters described in the above embodiments, etc. are not limited to those described in the above embodiments, etc., and may be other values, ranges, magnitude relationships, etc.

[0085] Additionally, in the above-described embodiment, the vehicle 1 is provided with a battery 12 as a power source. That is, in the above-described embodiment, the vehicle 1 is an electric vehicle (EV) or a hybrid electric vehicle (HEV), but the present invention is not limited to this example. That is, the vehicle 1 may be, for example, a gasoline-powered vehicle or a fuel-powered vehicle.

[0086] Furthermore, the series of processes described in the above embodiments may be performed by hardware (circuits), software (programs), or a combination of hardware and software. When performed by software, the software is composed of a group of programs for causing a computer to execute each function. Each program may be, for example, pre-installed in the computer, or may be installed on the computer from a network or recording medium.

[0087] Furthermore, the various examples described above may be applied in any combination.

[0088] The effects described in this specification are merely examples and are not limiting, and other effects may also be present.

[0089] The present disclosure may also be configured as follows: (1) A device for providing driving assistance for a target vehicle using abductive reasoning, comprising: an information acquisition unit configured to acquire observation information including one or more pieces of first event information; an abductive reasoning unit configured to perform the abductive reasoning by applying background knowledge information including a plurality of pieces of evidence information to the observation information, thereby inferring a plurality of hypotheses, each of which includes one or more pieces of second event information; a weight setting unit configured to individually set weights indicating importance for the evidence information included in the background knowledge information in accordance with the avoidance priority of the event indicated by the evidence information; and a determination unit configured to determine a maximum likelihood hypothesis, which is the hypothesis with the lowest total cost, from the plurality of hypotheses based on a cost for each piece of first event information included in the observation information and a cost for each piece of second event information included in each of the hypotheses. (2) The driving assistance device described in (1) above, wherein the weight setting unit is configured to set the weight in accordance with the level of urgency and the level of encounter frequency of the event, as the avoidance priority. (3) The driving assistance device according to (2) above, wherein the weight setting unit is configured to prioritize setting of the weight according to the level of urgency over setting of the weight according to the level of encounter frequency. (4) The driving assistance device according to (2) or (3) above, wherein the weight setting unit is configured to set the weight corresponding to the event having a relatively high encounter frequency relatively high. (5) The driving assistance device according to any of (2) to (4) above, wherein the weight setting unit is configured to set the weight corresponding to a immediately preceding event that is the event having a relatively high level of urgency relatively high. (6) The driving assistance device according to (5) above, wherein the weight setting unit is configured to set the weight corresponding to the immediately preceding event to a value equal to or greater than half of the total value of the weights in the background knowledge information. (7) The driving assistance device according to (5) or (6) above, wherein the weight setting unit is configured to set the weight corresponding to the immediately preceding event in accordance with at least one of an occurrence frequency and a margin time to occurrence of the immediately preceding event.(8) The driving assistance device according to any of (5) to (7) above, wherein the weight setting unit is configured to distribute and set weights of events other than the immediately preceding event (= the total value of the weights in the background knowledge information - the weight of the immediately preceding event) according to the encounter frequency of each of the other events. (9) The driving assistance device according to any of (1) to (8) above, wherein the weight setting unit is configured to periodically update the set value of the weight for each of the evidence information. (10) The driving assistance device according to any of (1) to (9) above, further comprising: a control unit configured to perform driving assistance for the target vehicle by using the maximum likelihood hypothesis. (11) A vehicle equipped with the driving assistance device according to any of (1) to (10) above. (12) A method for providing driving assistance for a target vehicle using abductive reasoning, the driving assistance method including: acquiring observation information including one or more pieces of first event information; applying background knowledge information including a plurality of pieces of evidence information to the observation information to perform the abductive reasoning, thereby inferring a plurality of hypotheses, each of which includes one or more pieces of second event information; individually setting weights indicating importance for the evidence information included in the background knowledge information, according to the avoidance priority of the event indicated by the evidence information; and determining a maximum likelihood hypothesis, which is the hypothesis with the lowest total cost, from the plurality of hypotheses, based on a cost for each piece of first event information included in the observation information and a cost for each piece of second event information included in each of the hypotheses.

[0090] The vehicle control unit 11 shown in FIGS. 1 and 2 can be implemented by circuitry including at least one semiconductor integrated circuit, such as at least one processor (e.g., a central processing unit (CPU)), at least one application-specific integrated circuit (ASIC), and / or at least one field-programmable gate array (FPGA). The at least one processor can be configured to perform all or some of the various functions of the vehicle control unit 11 shown in FIGS. 1 and 2 by reading instructions from at least one non-transitory, tangible computer-readable medium. Such medium can take various forms, including, but not limited to, various magnetic media such as hard disks, various optical media such as CDs or DVDs, and various semiconductor memories (i.e., semiconductor circuits) such as volatile or non-volatile memories. Volatile memories can include DRAM and SRAM. Non-volatile memories can include ROM and NVRAM. An ASIC is an integrated circuit (IC) specialized to perform all or some of the various functions of the vehicle control unit 11 shown in FIGS. 1 and 2. The FPGA is an integrated circuit that is designed to be configurable after manufacture so as to execute all or part of the various functions of the vehicle control unit 11 shown in FIGS.

Claims

1. A device that provides driving assistance for a target vehicle using abductive reasoning, comprising: an information acquisition unit configured to acquire observation information including one or more pieces of first event information; abductive reasoning unit configured to apply background knowledge information including multiple pieces of evidence information to the observation information to perform the abductive reasoning, thereby inferring multiple hypotheses, each of which includes one or more pieces of second event information; a weight setting unit configured to individually set weights indicating importance for the evidence information included in the background knowledge information, in accordance with the avoidance priority of the event indicated by the evidence information; and a determination unit configured to determine a maximum likelihood hypothesis from the multiple hypotheses, which is the hypothesis with the lowest total cost, based on the cost of each piece of first event information included in the observation information and the cost of each piece of second event information included in each of the hypotheses.

2. The driving assistance device according to claim 1, wherein the weight setting unit is configured to set the weight according to the degree of urgency and frequency of encounter of the event as the avoidance priority.

3. The driving assistance device according to claim 2, wherein the weight setting unit is configured to prioritize the setting of the weight according to the level of urgency over the setting of the weight according to the level of encounter frequency.

4. The driving assistance device according to claim 2, wherein the weight setting unit is configured to set the weight corresponding to the event with a relatively high encounter frequency to be relatively high.

5. A driving assistance device according to any one of claims 2 to 4, wherein the weight setting unit is configured to set the weight corresponding to the immediately preceding event, which is the event of relatively high urgency, relatively high.

6. The driving assistance device according to claim 5, wherein the weight setting unit is configured to set the weight corresponding to the immediately preceding event to a value equal to or greater than half of the total value of the weights in the background knowledge information.

7. The driving assistance device according to claim 5, wherein the weight setting unit is configured to set the weight corresponding to the immediately preceding event in accordance with at least one of the frequency of occurrence of the immediately preceding event and the time to occurrence of the immediately preceding event.

8. The driving assistance device according to claim 5, wherein the weight setting unit is configured to distribute and set the weight of events other than the immediately preceding event (= the sum of the weights in the background knowledge information - the weight of the immediately preceding event) according to the encounter frequency of each of the other events.

9. The driving assistance device according to any one of claims 1 to 4, wherein the weight setting unit is configured to periodically update the set value of the weight for each piece of evidence information.

10. A driving assistance device according to any one of claims 1 to 4, further comprising a control unit configured to provide driving assistance for the target vehicle using the maximum likelihood hypothesis.

11. A vehicle equipped with a driving assistance device according to any one of claims 1 to 4.

12. A method for providing driving assistance for a target vehicle using abductive reasoning, comprising: acquiring observation information including one or more pieces of first event information; applying background knowledge information including a plurality of pieces of evidence information to the observation information to perform the abductive reasoning, thereby inferring a plurality of hypotheses, each of which includes one or more pieces of second event information; individually setting weights indicating importance for the evidence information included in the background knowledge information, according to the avoidance priority of the event indicated by the evidence information; and determining a maximum likelihood hypothesis, which is the hypothesis with the lowest total cost, from among the plurality of hypotheses, based on the cost for each piece of first event information included in the observation information and the cost for each piece of second event information included in each of the hypotheses.

Images