JP2009051430A - Running support system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a running support system allowing a driver to recognize various risks existent in usual running environment and perform assist control to avoid the recognized risks in advance. <P>SOLUTION: This running support system makes a plan of the track where one's own vehicle runs to run in a place having the risk to the minimum extent by obtaining the distribution of risks expressed as a contour line by combining the information about the measured distance with the recognized risks and avoiding a place having high contour line (the place having high level of risk). When making this plan of the target track, how one's own vehicle behaves by control inputs is computed by using a vehicle model, and the target track within fixed time for suppressing the risk in front of one's own vehicle to the minimum extent is computed based on the computed vehicle behavior by using the intellectualizing technology. This running support system performs the assist control to enable one's own vehicle to run along the target track and avoid the recognized risk in advance. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a driving support system that assists a driver in accordance with a risk ahead of a vehicle.

  In general, when a vehicle such as an automobile is driven, the situation determination of a risk factor (risk) existing around the traveling direction depends on the driver's own attention and observation. Therefore, the risk may be overlooked depending on the physical condition and psychological state of the driver.

  In order to cope with such a problem, various techniques for detecting a dangerous state on the vehicle side and avoiding the danger have been developed.

  For example, in Patent Document 1, when there are a plurality of obstacles, an obstacle that is highly likely to collide with the host vehicle is selected with high accuracy, and the collision with the selected obstacle is accurately detected. Predictive techniques are disclosed.

  In Patent Document 2, it is determined whether or not an object on the road does not interfere with traveling even if paper or cloth collides, and unnecessary deceleration control is performed against the driver's intention. A technique for preventing a decrease in drivability due to continuous maintenance is disclosed.

Furthermore, in Patent Document 3, an obstacle to the vehicle is accurately determined in advance, and the vehicle behavior control device operates properly over the entire avoidance travel in consideration of various travel information. The technique which can perform appropriately is disclosed.
JP 2002-274344 A JP-A-9-11870 JP 2004-230947 A

  However, in the conventional techniques as disclosed in Patent Documents 1 to 3, the control is intervened only after the risk of collision becomes extremely high, so that the risk existing during normal traveling is avoided in advance. There is no special assist function.

  In addition, the conventional technology mainly deals with obstacles, and responds to risks that can be predicted in the actual driving environment such as people jumping out from the side of parked cars and cars jumping out from intersections. I can't.

  The present invention has been made in view of the above circumstances, and provides a driving support system capable of recognizing various risks existing in a normal driving environment and performing assist control so as to avoid the recognized risks. The purpose is that.

  In order to achieve the above object, a driving support system according to the present invention is a driving support system that supports a driver's operation according to a risk existing in the traveling direction of the host vehicle, and senses an external environment in the traveling direction of the host vehicle. A risk recognizing unit for recognizing a risk included in the external environment, a trajectory plan calculating unit for planning a course on which the host vehicle should travel as a target trajectory with less risk, and calculating the target trajectory sequentially, An assist determination unit that determines the type of assist control according to the risk of the target locus, and an assist control unit that determines an assist control amount according to the type of the assist control and performs assist control on the target locus. It is characterized by.

  According to the present invention, it is possible to recognize various risks existing in a normal driving environment, perform assist control so as to avoid the recognized risks, and improve preventive safety.

  Embodiments of the present invention will be described below with reference to the drawings. 1 to 11 relate to an embodiment of the present invention, FIG. 1 is a configuration diagram of a driving support system, FIG. 2 is an explanatory diagram showing a concept of risk, and FIG. 3 is an explanatory diagram showing a risk distribution situation and a target locus. 4 is an explanatory diagram showing the setting of the target trajectory point, FIG. 5 is an explanatory diagram showing the trajectory control, FIG. 6 is a flowchart showing the main routine of the driving support control, FIG. 7 is a flowchart of the target trajectory calculation routine, and FIG. FIG. 9 is an explanatory diagram of a two-wheel model, FIG. 10 is a flowchart of a trajectory calculation routine, and FIG. 11 is an explanatory diagram of trajectory calculation.

  As shown in FIG. 1, the driving support system 1 according to the present embodiment performs driving assistance for a driver based on risk recognition and a minimum risk trajectory plan, and includes a plurality of sensors and functional units of the system. Are connected to each other via the information communication bus 2. The sensor connected to the information communication bus 2 is a vehicle information sensor comprising a plurality of sensors for detecting vehicle speed, brake operation (hydraulic pressure), steering angle, accelerator opening, turn signal switch operation information, light switch information, acceleration, and the like. Various information detected by the vehicle information sensor group 3 is transmitted to a plurality of devices via the bus 2 and shared.

  Devices connected to the bus 2 include a risk recognition device 4 that recognizes risks contained in the external environment around the vehicle from images, radar, etc., and grasps a risk distribution in the environment outside the vehicle, a GPS or stereo image recognition device, etc. Self-vehicle position measurement device 5 that recognizes the relative position of the vehicle based on information, a trajectory plan calculation device 6 that calculates a trajectory within a certain time to minimize the risk ahead of the vehicle, and a risk minimum trajectory An assist determination device 7 that performs an assist determination based on the control amount for determining, and an assist control device 8 that determines the assist amount based on the assist determination and controls the control device group 9. The control device group 9 is a device group including actuators such as an electric power steering, an electric throttle, and a brake actuator, an actuator for controlling left and right torque distribution under an emergency avoidance state, and the like.

  The driving assistance by the driving support system 1 is usually led by a veteran driver or the like so as not to approach a high-risk place. Avoidance control. A high-risk place refers to a place where there is no obstacle but a high risk, such as a person jumping out from the side of a parked car or a car jumping out of an intersection.

  In the present embodiment, the risk recognized by the risk recognition device 4 is recognized from the feature amount of the front image captured by the in-vehicle camera. The recognition of the risk from this image is performed by, for example, the online risk learning system technology proposed in Japanese Patent Application No. 2007-77625 by the applicant or the image processing technology of a tree structure proposed in Japanese Patent Application Laid-Open No. 2006-178857. Can be used.

  In particular, the risk recognition technique detailed in Japanese Patent Application No. 2007-77625 is the most suitable technique for this system, and an image feature amount of an N-dimensional vector (according to an event such as an abrupt accelerator return operation or a brake depression) ( Edge information, motion information, color information, etc.) is converted into a one-dimensional state, and the risk contained in the external environment is determined based on the correlation between this state and teacher information created from vehicle information (driver operation information). Learning and recognition.

  As shown in Fig. 2, the risk recognized here means that the risk is higher for people than for cars and road structures, and there are also high risks in the vicinity. Even if there is no risk, it will respond to the prediction risk that can be expected. FIG. 2 shows the concept of risk recognized by the system. The area where the risk exists is surrounded by a frame, and the higher the risk, the higher the concentration. Risk areas exist not only in areas that are visible to people, such as pedestrians, cars, curbs, road signs, and pedestrian crossings, but also in areas that are not visible to people, such as behind cars and at the edges of pedestrian crossings. This indicates that the system is aware of the risks.

  Further, the risk recognizing device 4 obtains a risk distribution as shown in FIG. 3 by combining distance measurement information with a stereo camera, a millimeter wave radar, or the like in order to obtain position information of the recognized risk in three dimensions. This risk distribution situation can be expressed as a contour line, and the trajectory plan calculation device 6 avoids a place with a high contour line (a place with a high risk level) and travels in a place with the lowest risk. Planning the trajectory to be run.

  The trajectory plan calculation apparatus 6 forms the central function of the present system, and includes a vehicle behavior calculation unit 10 in detail. The vehicle behavior calculation unit 10 calculates what behavior is to be performed by a control input using a vehicle model, and uses intelligent technology to calculate a target trajectory within a certain period of time that minimizes forward risk from the calculated vehicle behavior. calculate. The vehicle model is adaptively updated so as to maintain an optimum state according to the environment (road surface, occupant, tire, etc.).

  The target trajectory obtained here has two types of trajectories, a first target trajectory and a second target trajectory, as will be described next.

  The first target trajectory is a trajectory that minimizes risk by assuming forced control of the accelerator, brake, and steering. This first target trajectory is planned with respect to the limit of the safety line, and if the risk value finally exceeds a certain level, there is a high possibility that an accident will occur, so it is forced to avoid the current minimum risk Used for.

  In the present embodiment, the first target trajectory is planned assuming an assist operation to the accelerator, the brake, and the steering, but may be planned assuming an assist operation to the brake and the steering. good.

  The second target trajectory is a trajectory that minimizes risk by assuming only control in the lateral direction (vehicle width direction) by steering while maintaining the current speed. This second target trajectory is usually planned to assist the driver while driving and is used to assist the steering along that trajectory so that the current flow is maintained and risk is minimized. .

  However, if the risk value of the second target trajectory exceeds a certain level, there is a high possibility of an accident, so assistance including deceleration is performed. If it is still dangerous, the risk value in the first target trajectory increases, and the avoidance action is forcibly entered.

  The trajectory plans of the first target trajectory and the second target trajectory are optimized in a learning manner using a genetic algorithm (GA). Next, a method for sequentially obtaining each trajectory plan using GA will be described.

First, as shown in FIG. 4, a coordinate system fixed on the ground is set, and the vehicle position (target trajectory point) for each unit time t is expressed by relative position change amounts dx and dy. It is defined as gene dna as shown in the formula. Then, a simulation is performed on the target trajectory for each generation using the control model and the vehicle model, and optimization by GA is performed using the evaluation result.
dna = [dx _t , dy _t , dx _{t + 1} , dy _{t + 1} ,..., dx _{t + n} , dy _{t + n} ] (1)

  The generation of the initial individual for each optimization cycle by GA can be optimized at a higher speed by setting based on the previously generated target locus. Further, the unit time of the trajectory may be changed based on the vehicle speed at that time, and optimization according to the situation may be performed.

  As the control model used for the simulation of the target trajectory, different models are used when the first target trajectory is calculated and when the second target is calculated. The control model used when calculating the first target trajectory is a model of control that finally performs automatic avoidance in the case of danger, and controls the steering, accelerator, and brake to follow the target trajectory as much as possible. On the other hand, the control model used when calculating the second target locus is a model that controls only the lateral direction, that is, the steering.

  The vehicle motion is calculated through these control models, and the position / velocity that will actually pass is obtained. The vehicle model used for the calculation of the vehicle motion at this time is adaptively changed with respect to the external environment (for example, road surface friction coefficient, load capacity, etc.) using a two-wheel model or a high-dimensional model, and the vehicle behavior at that time is changed. By making it reproducible, the accuracy of the system can be improved.

Evaluation in the optimization process by GA is performed using an evaluation function E shown in the following equation (2), and optimization is performed so that the value of the evaluation function E is minimized. The evaluation function E in the equation (2) is expressed by a value obtained by adding the integrated value of the risk index obtained by multiplying the risk distribution value of each point i by the speed coefficient by the simulation and the integrated value of the assist control amount.
_{E = Σ (w 1 R i} V i + w 2 Ast i + w 3 Abrk i + w 4 Aacc i + w 5 Navi i + w 6 Posi i) ... (2)
However, R: Risk (per simulation discrete time)
Ast: Steering control amount
Abrk: Brake control amount
Aacc: accelerator control amount
Navi: error between the route guidance of the navigation device and the target trajectory
Posi: Position error
w _{1 to} w ₆ : weighting factors

  Further, the target locus planning time (how far ahead is planned) is made variable according to the speed and the forward risk monitoring status. As described above, the first target trajectory is a trajectory that will be least risky at the time t. Therefore, if the evaluation value at this time is equal to or greater than a certain value, there is a high possibility that an accident will occur. Take evasive action to minimize the risk of intervention.

  The target trajectory (ideal trajectory) optimized as described above is referred to by the assist determination device 7, and it is determined whether or not an avoidance operation is required, and the type of assist control is determined. Basically, when the evaluation value of the first target locus is equal to or greater than the first reference value, the assist determination device 7 instructs the assist control device 8 to perform forced avoidance control. When the first target locus is less than the first reference value and the evaluation value of the second target locus is greater than or equal to the second reference value, the assist determination device 7 performs steering assist and deceleration assist along the second target locus. The assist control device 8 is instructed. On the other hand, when the evaluation value of the second target locus is less than the second reference value, it is safe, so the assist determination device 7 instructs the assist control device 8 to perform steering assist along the second target locus.

  In the assist state, the assist control device 8 makes the driver easy to operate in that direction and performs a slight increase control by reducing the direction in which the steering reaction force is to be induced and increasing the reverse side. The deceleration assist assists by making the engine brake easier to use by making the accelerator pedal less responsive. Further, the avoidance performance is increased in advance by increasing the gain of the brake, the left / right torque control, etc. according to the degree of risk (evaluation value). During forced avoidance control, trajectory control based on the vehicle position and future prediction deviation is performed so as to match the target trajectory, and forced avoidance is performed.

For example, as shown in FIG. 5, when the deviation between the target course point (target trajectory point) and the vehicle position at time t is Ex, Ey in a coordinate system fixed to the vehicle, the vehicle speed V (Vx, Vy) Deviations Ex + δt and Ey + δt after δt seconds are predicted. Then, using the predicted deviations Ex + δt, Ey + δt, the steering, the accelerator, and the brake are controlled by the control amounts as shown in the following equations (3) and (4), and the time δt from the time t at the yaw angle θ and the speed Vm. The traveling of the host vehicle is controlled toward the target course point leading to.
Ast = Kp (Ey + δt) + Ki (Ey / s) (3)
Aacc & Abrk = Kp2 (Ex + δt) + Ki2 (Ex / s) (4)

  Note that Kp and Ki in the expression (3) are steering control gains. Kp2 and Ki2 in equation (4) are accelerator or brake control gains, which are decomposed into an accelerator when the deviation is + (plus) and into a brake when the deviation is-(minus).

  Next, the processing of the system will be described with reference to the flowcharts shown in FIGS.

  The flowchart of FIG. 6 shows the main routine of the overall driving support control. The first steps S1 and S2 of this main routine are processes performed by the risk recognition device 4. First, in step S1, environmental risks are recognized from the image captured by the in-vehicle camera, and a risk distribution is created in step S2.

  Information on the risk distribution created in step S2 is transmitted to the trajectory plan calculation device 6 via the bus 2. The trajectory plan calculation device 6 acquires various vehicle information including risk distribution via the bus 2 in step S3, and executes target trajectory (first target trajectory, second target trajectory) calculation processing in step S4. This target trajectory is calculated by the routine shown in the flowchart of FIG. 7, and details will be described later.

  Once the target trajectory is calculated, the assist determination device 7 adjusts the vehicle gain from the evaluation value (risk degree) of the target course, and performs a determination process of avoidance control or assist control. That is, in step S5, the evaluation value E1 of the first target locus is compared with the gain adjustment threshold value KHS. As a result, if E1 ≦ KHS, the vehicle control gain is set to a normal gain in step S6, and if E1> KHS, the vehicle control gain is set to a higher Hi gain than normal in step S7.

  Then, after setting the vehicle control gain, it is determined in step S8 whether or not an avoidance operation is necessary by comparing the evaluation value E1 with the avoidance determination threshold value (first reference value) SH1. As a result, if E1 ≧ SH1, it is determined that avoidance is impossible, and avoidance control is instructed to the assist control device 8 in step S12, and avoidance control such as pre-crash control is executed by the assist control device 8 to perform one cycle of processing. finish.

  On the other hand, if E1 <SH1, it is determined that avoidance is possible. In step S9, the evaluation value E2 of the second target locus is compared with the assist determination threshold value (second reference value) SH2, and assist or steering with deceleration is performed. Judge whether only assist. If E2 <SH2, it is determined that the target course can be safely maintained with the assistance of only the steering operation, and the steering control is instructed to the assist control device 8, and the assist control device 8 causes the second target locus in step S10. Steering assist along is performed, and one cycle of processing is terminated. If E2 ≧ SH2, it is determined that deceleration is necessary, and the assist control device 8 is instructed to perform brake control and steering control. In step S11, the assist control device 8 performs steering assist with brake control. Control to the second target locus is performed, and one cycle of processing is terminated.

  Next, the target locus (first target locus, second target locus) calculation process in step S4 of the above main routine will be described.

  The target trajectory is calculated by a target trajectory calculation routine shown in FIG. 7. First, risk / vehicle information is acquired in the first step S21, and model parameters such as road surface friction coefficient and vehicle weight in the vehicle model are adaptively applied in step S2. Update to Specifically, the parameters of the vehicle model are updated by a vehicle model update routine shown in FIG.

  Here, the vehicle model update routine of FIG. 8 will be described. Updating the vehicle model basically simulates the trajectory from the yaw rate and speed, and optimizes the model so that the sensor value of the actually driven yaw rate matches the model value calculated by putting the same input into the model. Turn into.

  That is, first, in step S31, the model yaw rate yrs_md is calculated from the operation value one step before, and in step S32, the current yaw rate (sensor value) yrs_real is obtained. In step S33, the absolute value | yrs_md-yrs_real | of the difference between the model yaw rate yrs_md and the current yaw rate yrs_real is compared with the threshold value SHyrs. If | yrs_md-yrs_real |

  On the other hand, if | yrs_md-yrs_real |> SHyrs, the road surface friction coefficient and the vehicle weight are optimized in step S34, the updated parameters are reflected in the model in step S34, and the process ends.

  For example, when performing a lateral simulation with a two-wheel model, replace the change in the friction coefficient of the road surface with the characteristic change of the cornering power value and estimate it by back calculation, or change the cornering power value slightly and adaptively use the change gradient Optimize to. As for the weight parameter, the vehicle weight sensor value is directly reflected, or the model response change due to the weight change is replaced with the cornering power change, and the cornering power is optimized to match the actual behavior.

That is, as shown in FIG. 9, when performing a lateral simulation with a two-wheel model, as shown in the following equation (5), the equation of motion of the lateral translational motion is the cornering forces Ff and Fr of the front and rear wheels. The vehicle body mass m and the lateral acceleration d ² y / dt ² can be expressed.
m · d ² y / dt ² = 2 · Ff + 2 · Fr (5)

On the other hand, the equation of motion of rotation around the center of gravity can be expressed by the following equation (6) by the distances lf and lr from the center of gravity to the front and rear axles, the yawing inertia moment Iz of the vehicle body, and the yaw angle θ.
Iz · (d ² θ / dt ² ) = 2 · Ff · lf−2 · Fr · lr (6)

When the vehicle speed V and the translational speed (side slip speed) in the lateral direction of the center of gravity are used, the lateral acceleration d ² y / dt ² is expressed by the following equation (7).
d ² y / dt ² = V · (dβ / dt + dθ / dt) (7)

The cornering force responds close to the first-order lag with respect to the side slip angle of the tire. If this lag is ignored, the following (8) depends on the cornering powers Kf and Kr of the front and rear wheels and the slip angles βf and βr of the front and rear wheels. , (9).
Ff = −Kf · βf (8)
Fr = −Kr · βr (9)

When equivalent cornering power is used in consideration of the influence of rolls and suspension in the cornering power, the slip angles βf and βr of the front and rear wheels are simply expressed by the following equations (10) and (11) depending on the actual steering angle δ. Can be
βf = β + (dθ / dt · lr) / V−δ (10)
βr = β− (dθ / dt · lf) / V (11)

  The above equations (5) to (11) are basic equations of motion. These equations of motion are represented by state variable expressions, and various parameters are estimated by developing parameter control rules and developing adaptive control theory. be able to. Then, the cornering power of the actual vehicle is obtained from the estimated parameters, and the road surface friction coefficient in the non-linear region can be estimated by comparing with the value of the high friction coefficient road for each front and rear wheel, for example.

  In addition, as represented by the “vehicle motion model generation device” of Japanese Patent Application Laid-Open No. 2004-249812 previously proposed by the present applicant, the vehicle model itself is constructed by a neural network and adaptively changed. Anyway.

  After the vehicle model is updated, the process proceeds to step S23 of the target trajectory calculation routine, and the trajectory plan time is dynamically changed and set according to the state of the external environment. For example, the trajectory planning time is determined by the average and maximum of the risk in the monitoring area, and when the average and maximum of the risk is less than a certain value, the trajectory planning time is shortened as the reference time x 0.5, and when it is more than a certain value The reference time is extended to 1.5.

  When the trajectory plan time is set, the first target trajectory and the second target trajectory are calculated using GA in steps S23 and S24, respectively. The calculation of the first target trajectory and the second target trajectory is basically the same as the model used for the respective trajectory simulations, except that the evaluation values for the simulation results are different, and is represented by the flowchart shown in FIG. explain.

The trajectory calculation routine shown in FIG. 10 performs trajectory simulation in step S42 when initial individuals are generated (for example, 10) by using the target trajectory points at time t as genes in the first step S41. The trajectory point (X, Y) of the vehicle can be calculated from the vehicle body slip angle β, yaw angle θ, and speed V according to the following equations (12) and (13), as shown in FIG.
X = X0 + V∫cos (β + θ) dt (12)
Y = Y0 + V∫sin (β + θ) dt (13)

  After executing the trajectory simulation in step S42, an evaluation value is calculated in step S43 in order to evaluate the simulation result. This evaluation value calculation is the calculation of the evaluation value E according to the above-described equation (2). In the case of the first target locus, the evaluation value E1 is calculated, and in the case of the second target locus, the evaluation value E2 is calculated. .

  Then, the evaluation value calculation based on the above trajectory simulation is repeated until the simulation time in step S44 ends, and it is determined in step S44 whether or not the simulation of all individuals has been completed. Until the simulation of all individuals is completed, after the simulation of each individual, the individual number is updated in step S46 to continue the trajectory simulation and the evaluation value calculation. When the simulation of all individuals is completed, the next generation is performed in step S47. In order to rank and select the individuals that should survive, the individuals are sorted.

  Then, in the next step S48, a next generation individual is generated by crossover processing, and the evaluation value is recalculated. In step S49, the evaluation change rate is compared with the threshold value EHS, and optimization convergence determination is performed. When the evaluation change rate exceeds the threshold value EHS, the process returns to step S42, and the trajectory simulation, evaluation, and generation change are repeated. On the other hand, when the evaluation change rate has converged below the threshold value EHS, the trajectory at that time is output as an optimized target trajectory, and this calculation process is terminated.

  As described above, the driving support system according to the present embodiment enhances the preventive safety effect by assisting the steering, the brake, and the accelerator so that the vehicle travels on the route with the least risk according to the forward risk, In an emergency, forced avoidance control can be performed to improve preventive safety.

  Further, adaptively optimizing the vehicle model in response to changes in the external environment can improve the accuracy of the trajectory plan and improve the safety. Furthermore, by performing assist control according to the degree of risk, it is possible to avoid the risk in advance without causing the driver to feel uncomfortable.

Configuration diagram of the driving support system Same as above, explanatory diagram showing the concept of risk Same as above, explanatory diagram showing risk distribution status and target trajectory As above, an explanatory diagram showing the setting of the target locus point Same as above, explanatory diagram showing trajectory control Same as above, flowchart showing main routine of driving support control Same as above, flowchart of target locus calculation routine Same as above, flowchart of vehicle model update routine Same as above, illustration of two-wheel model Same as above, flowchart of locus calculation routine Same as above, illustration of locus calculation

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Driving assistance system 3 Vehicle information sensor group 4 Risk recognition apparatus 6 Trajectory plan calculation apparatus 7 Assist judgment apparatus 8 Assist control apparatus 9 Control device group 10 Vehicle behavior calculation part E1, E2 Evaluation value

Claims (15)

A driving support system that supports the operation of the driver according to the risk existing in the traveling direction of the host vehicle,
A risk recognition unit that senses the external environment in the direction of travel of the vehicle and recognizes the risk included in the external environment;
A trajectory plan calculation unit for planning a course on which the host vehicle should travel as a target trajectory with less risk, and calculating the target trajectory sequentially;
An assist determination unit that determines the type of assist control according to the risk of the target trajectory;
A driving support system, comprising: an assist control unit that determines an assist control amount according to a type of the assist control and performs assist control on the target locus.   As the target trajectory, two target trajectories are planned: a first target trajectory assuming an assist operation to an accelerator, a brake, and a steering, and a second target trajectory assuming only an assist operation to the steering, The driving support system according to claim 1.   The target trajectory is planned as two target trajectories: a first target trajectory assuming an assist operation to the brake and the steering and a second target trajectory assuming only an assist operation to the steering. Item 2. The driving support system according to item 1.   The driving support system according to claim 1, wherein the assist amount is increased in accordance with the planned risk degree of the target locus.   The driving support system according to claim 1, wherein the avoidance control is forcibly performed when the risk of the target locus is a certain level or more.   The driving support system according to claim 1, wherein a vehicle model is used for planning the target locus, and the vehicle locus is adaptively optimized to plan the target locus.   The driving support system according to claim 6, wherein the vehicle model includes a road surface parameter.   The driving support system according to claim 1, wherein the target locus is planned based on route information by a navigation device, and a locus with less risk is calculated along the route.   The driving support system according to claim 1, wherein the trajectory is optimized in a learning manner using a genetic algorithm in the planning of the target trajectory.   10. The driving support system according to claim 9, wherein an integrated value of risk distribution values through which the target locus passes is used as an evaluation value for optimization by the genetic algorithm.   11. The driving support system according to claim 10, wherein the evaluation is performed using a value obtained by multiplying the evaluation value by a speed coefficient.   The driving support system according to claim 9, wherein an integrated value of the assist control amount is used as an evaluation value of optimization by the genetic algorithm.   The driving support system according to claim 9, wherein a gene in the genetic algorithm is defined by a relative position of the host vehicle per unit time.   14. The driving support system according to claim 13, wherein the unit time is variable depending on the vehicle speed.   2. The driving support system according to claim 1, wherein the vehicle control gain is changed in advance corresponding to the risk of the external environment prior to the assist control to the target locus.

Images