System and method for generating emergency collision avoidance strategy for a vehicle

1 . A system for establishing a strategy for allowing a vehicle to avoid collision risk, the system being implemented by a processor, the system comprising:

a risk degree determination unit configured to determine a degree of collision risk with a surrounding object of an ego vehicle;

a vehicle trajectory analysis unit configured to analyze a future vehicle trajectory of the surrounding object;

an avoidance trajectory analysis unit configured to analyze a future avoidance trajectory of the ego vehicle corresponding to a plurality of avoidance strategies for avoiding collision with the surrounding object; and

a collision avoidance strategy determination unit configured to learn the future vehicle trajectory of the surrounding object and the future avoidance trajectory of the ego vehicle and output an avoidance strategy selected from the plurality of avoidance strategies,

wherein the collision avoidance strategy determination unit is configured to output the avoidance strategy with a set high priority when it is determined that multiple avoidance strategies among the plurality of avoidance strategies are possible as a result of learning.

2 . The system of claim 1 , further comprising:

a bird's eye view generation unit configured to generate a bird's eye view image reflecting the degree of collision risk, a result of analyzing the future vehicle trajectory of the surrounding object, and a result of analyzing the future avoidance trajectory of the ego vehicle.

3 . The system of claim 2 , wherein the collision avoidance strategy determination unit is configured to:

learn a stacked bird's eye view of a plurality of bird's eye view images reflecting a plurality of avoidance trajectory analysis results; and

output the avoidance strategy selected from the avoidance strategies through a convolutional neural network (CNN) technique.

4 . The system of claim 3 , further comprising:

a surrounding object information detection unit configured to acquire state information including a location, a speed, and an acceleration of the surrounding object through a sensor mounted on the vehicle; and

a lane information detection unit configured to acquire lanes within a predetermined distance from the ego vehicle and shoulder information through the sensor mounted on the vehicle or a precision map.

5 . The system of claim 4 , wherein the risk degree determination unit is configured to:

calculate a longitudinal collision risk index, a time to lane crossing (TLC), and a lateral collision risk index using the state information and lane and shoulder information; and

determine the degree of collision risk.

6 . The system of claim 4 , wherein the vehicle trajectory analysis unit is configured to analyze the future vehicle trajectory of the surrounding object through an estimation algorithm or a deep-learning algorithm using the state information and the lane and shoulder information.

7 . The system of claim 4 , wherein the bird's eye view generation unit configured to generate the bird's eye view image reflecting a past traveling trajectory, a current location, a future vehicle trajectory, and a final location of the surrounding object, a current location and an avoidance trajectory prediction result of the ego vehicle, and lane and shoulder information within a predetermined distance from the ego vehicle.

8 . The system of claim 7 , wherein the bird's eye view generation unit is configured to generate the bird's eye view image to be displayed separately according to the degree of collision risk.

9 . The system of claim 4 , wherein the plurality of avoidance strategies includes a lane change to left (LCL), a lane change to right (LCR), a steering to shoulder stop (SSS), an evasive steering to left (ESL), and an evasive steering to right (ESR).

10 . The system of claim 4 , wherein the collision avoidance strategy determination unit is configured to output a result value corresponding to safety when it is analyzed that no collision with the surrounding object occurs as a result of learning without considering the future avoidance trajectory of the ego vehicle.

11 . The system of claim 4 , wherein the collision avoidance strategy determination unit is configured to output a result value corresponding to collision when it is analyzed that it is impossible to avoid collision with the surrounding object as a result of learning the plurality of avoidance strategies.

12 . A method of establishing a strategy for allowing a vehicle to avoid collision risk, the method comprising:

acquiring state information including a location, a speed, and an acceleration of a surrounding object and lane and shoulder information within a predetermined distance from an ego vehicle;

determining a degree of collision risk with the surrounding object of the ego vehicle by using the state information and the lane and shoulder information;

analyzing a future vehicle trajectory of the surrounding object by using the state information and the lane and shoulder information;

analyzing a future avoidance trajectory of the ego vehicle corresponding to a plurality of avoidance strategies for avoiding collision with the surrounding object;

learning the future vehicle trajectory of the surrounding object and the future avoidance trajectory of the ego vehicle; and

outputting an avoidance strategy selected from the plurality of avoidance strategies,

wherein outputting the avoidance strategy is configured to output the avoidance strategy with a set high priority when it is determined that multiple avoidance strategies among the plurality of avoidance strategies are possible as a result of learning.

13 . The method of claim 12 , further comprising:

generating a bird's eye view image reflecting results of determining the degree of collision risk, analyzing the future vehicle trajectory of the surrounding object, and analyzing the future avoidance trajectory of the ego vehicle,

wherein outputting the avoidance strategy includes:

learning a stacked bird's eye view of a plurality of bird's eye view images reflecting a plurality of avoidance trajectory analysis results, and

outputting the avoidance strategy selected from the avoidance strategies through a convolutional neural network (CNN) technique.

14 . The method of claim 13 , wherein generating the bird's eye view image includes generating the bird's eye view image reflecting a past traveling trajectory, a current location, a future vehicle trajectory, and a final location of the surrounding object, a current location and an avoidance trajectory analysis result of the ego vehicle, and the lane and shoulder information within a predetermined distance from the ego vehicle; and

separately displaying the bird's eye view image according to the degree of collision risk.

15 . The method of claim 13 , wherein the plurality of avoidance strategies includes a lane change to left (LCL), a lane change to right (LCR), a steering to shoulder stop (SSS), an evasive steering to left (ESL), and an evasive steering to right (ESR).

16 . The method of claim 15 , wherein outputting the avoidance strategy includes:

determining whether the surrounding object is present at a left side of the ego vehicle;

determining whether a shoulder is present within a predetermined distance from the ego vehicle when the surrounding object is present at the left side of the ego vehicle;

determining whether collision occurs when an SSS strategy is applied when the shoulder is present;

outputting the SSS strategy when the collision does not occur when the SSS strategy is applied; and

outputting a result value corresponding to the collision when the collision occurs.

17 . The method of claim 16 , further comprising:

determining whether the collision occurs when an ESR strategy is applied when no shoulder is present by determining whether the shoulder is present;

outputting the ESR strategy when no collision occurs when the ESR strategy is applied;

determining whether the collision occurs when an LCR strategy is applied when the collision occurs when the ESR strategy is applied;

outputting the LCR strategy when no collision occurs when the LCR strategy is applied; and

outputting a result value corresponding to the collision when the collision occurs.

18 . The method of claim 16 , further comprising:

determining whether the collision occurs when an ESL strategy is applied when no surrounding object is present at the left side of the ego vehicle by the determining of whether the surrounding object is present at the left side of the ego vehicle;

outputting the ESL strategy when no collision occurs when the ESL strategy is applied;

determining whether the collision occurs when an LCL strategy is applied when the collision occurs when the ESL strategy is applied;

outputting the LCL strategy when no collision occurs when the LCL strategy is applied; and

outputting a result value corresponding to the collision when the collision occurs.

19 . The method of claim 15 , wherein outputting the avoidance strategy includes outputting a result value corresponding to safety when it is analyzed that no collision with the surrounding object occurs as a result of learning without considering the future avoidance trajectory of the ego vehicle.