Methods and systems for smooth trajectory generation for a self-driving vehicle
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
B60W-030/10
B62D-015/02
출원번호
US-0616813
(2012-09-14)
등록번호
US-9120485
(2015-09-01)
발명자
/ 주소
Dolgov, Dmitri
출원인 / 주소
Google Inc.
대리인 / 주소
McDonnell Boehnen Hulbert & Berghoff LLP
인용정보
피인용 횟수 :
24인용 특허 :
6
초록▼
A vehicle configured to operate in an autonomous mode is provided. The vehicle is configured to follow a baseline trajectory. Changes to the baseline trajectory are received by a computer system associated with the vehicle. The computer system creates a function representing the current trajectory o
A vehicle configured to operate in an autonomous mode is provided. The vehicle is configured to follow a baseline trajectory. Changes to the baseline trajectory are received by a computer system associated with the vehicle. The computer system creates a function representing the current trajectory of the vehicle, as well as one or more functions defining any desired trajectory criteria, and generates an optimization problem from the functions, which, when solved, provide a new trajectory for the vehicle to follow that moves efficiently and smoothly toward the changed baseline trajectory.
대표청구항▼
1. A method, comprising: controlling, using a computer system, a vehicle in an autonomous mode to follow a first trajectory, the first trajectory comprising a pre-determined path;receiving data indicating a second trajectory;defining a plurality of vertices based on the first trajectory;defining a f
1. A method, comprising: controlling, using a computer system, a vehicle in an autonomous mode to follow a first trajectory, the first trajectory comprising a pre-determined path;receiving data indicating a second trajectory;defining a plurality of vertices based on the first trajectory;defining a first set of cost functions for the plurality of vertices, wherein each cost function in the first set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to the second trajectory;defining a second set of cost functions for the plurality of vertices, wherein each cost function in the second set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to an estimated center of a road lane on which the vehicle is currently traveling;optimizing the location of the vertices in the plurality of vertices based on at least the costs assigned by the first set of cost functions and the second set of cost functions;determining a transitional trajectory based on the optimized vertices, wherein the transitional trajectory transitions from the first trajectory to the second trajectory; andcontrolling the vehicle in the autonomous mode to follow the transitional trajectory. 2. The method of claim 1, wherein an indication of the estimated center of the road lane is received from at least one sensor. 3. The method of claim 1, wherein each cost function in the second set assigns to a respective vertex in the plurality of vertices a respective cost further based on a sum of squares of curvature for the vertex and is configured to penalize a non-smooth trajectory. 4. The method of claim 1, further comprising: defining a third set of cost functions for the plurality of vertices, wherein each cost function in the third set assigns to a respective vertex in the plurality of vertices a respective cost based on at least one other criterion; andoptimizing the vertices in the plurality of vertices based on at least the costs assigned by the first set of cost functions, the second set of cost functions, and the third set of cost functions. 5. The method of claim 1, wherein optimizing comprises generating a non-linear optimization problem. 6. The method of claim 5, wherein the non-linear optimization problem is a gradient descent algorithm. 7. The method of claim 5, wherein the non-linear optimization problem is a conjugate descent algorithm. 8. The method of claim 5, wherein the non-linear optimization problem comprises computing a gradient as a function of a current state of the vertex. 9. The method of claim 1, wherein the plurality of vertices form a piecewise linear parametric curve. 10. A vehicle, comprising: a computer system configured to control the vehicle in an autonomous mode to follow a first trajectory, the first trajectory comprising a pre-determined path, receive data indicating a second trajectory, define a plurality of vertices based on the first trajectory, define a first set of cost functions for the plurality of vertices, wherein each cost function in the first set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to the second trajectory, define a second set of cost functions for the plurality of vertices, wherein each cost function in the second set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to an estimated center of a road lane on which the vehicle is traveling, optimize the vertices in the plurality of vertices based on at least the costs assigned by the first set of cost functions and the second set of cost functions, determine a transitional trajectory based on the optimized vertices, wherein the transitional trajectory transitions from the first trajectory to the second trajectory, and control the vehicle in the autonomous mode to follow the transitional trajectory. 11. The vehicle of claim 10, wherein an indication of the estimated center of the road lane is received from at least one sensor. 12. The vehicle of claim 10, wherein each cost function in the second set assigns to a respective vertex in the plurality of vertices a respective cost further based on a sum of squares of curvature for the vertex and is configured to penalize a non-smooth trajectory. 13. The vehicle of claim 10, wherein the computer system is further configured to generate a non-linear optimization problem. 14. The vehicle of claim 13, wherein the non-linear optimization problem is a gradient descent algorithm. 15. The vehicle of claim 13, wherein the non-linear optimization problem is a conjugate descent algorithm. 16. The vehicle of claim 13, wherein the non-linear optimization problem comprises computing a gradient as a function of a current state of the vertex. 17. A non-transitory computer readable medium having stored therein instructions executable by a computer system to cause the computer system to perform functions, the functions comprising: controlling a vehicle in an autonomous mode to follow a first trajectory, the first trajectory comprising a pre-determined path;receiving data indicating a second trajectory;defining a plurality of vertices based on the first trajectory;defining a first set of cost functions for the plurality of vertices, wherein each cost function in the first set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to the second trajectory;defining a second set of cost functions for the plurality of vertices, wherein each cost function in the second set assigns to a respective vertex in the plurality of vertices a respective cost based on a distance from the respective vertex to an estimated center of a road lane on which the vehicle is currently traveling;optimizing the vertices in the plurality of vertices based on at least the costs assigned by the first set of cost functions and the second set of cost functions;determining a transitional trajectory based on the optimized vertices, wherein the transitional trajectory transitions from the first trajectory to the second trajectory; andcontrolling the vehicle in the autonomous mode to follow the transitional trajectory. 18. The non-transitory computer readable medium of claim 17 wherein the instructions are executable to define a third set of cost functions for the plurality of vertices, wherein each cost function in the third set assigns to a respective vertex in the plurality of vertices a respective cost based on at least one other criterion and to optimize the vertices in the plurality of vertices based on at least the costs assigned by the first set of cost functions, the second set of cost functions, and the third set of cost functions.
Hoffmann, Walter; Papiernik, Wolfgang; Sauer, Tomas, Method for the determination of a rough trajectory to be followed in a positionally guided manner.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.