초록 ▼

An autonomous vehicle configured for active sensing may also be configured to weigh expected information gains from active-sensing actions against risk costs associated with the active-sensing actions. An example method involves: (a) receiving information from one or more sensors of an autonomous ve

An autonomous vehicle configured for active sensing may also be configured to weigh expected information gains from active-sensing actions against risk costs associated with the active-sensing actions. An example method involves: (a) receiving information from one or more sensors of an autonomous vehicle, wherein one or more control processes for an autonomous vehicle are based upon the information, (b) determining an information-improvement expectation that corresponds to an active-sensing action, (c) determining a risk-cost that corresponds to the active-sensing action; and (d) based on both (i) the information-improvement expectation for the active-sensing action and (ii) the risk-cost for the active-sensing action, determining whether the active-sensing action is advisable.

대표청구항 ▼

1. A computer-implemented method comprising: receiving information from one or more sensors of an autonomous vehicle, wherein one or more control processes for the autonomous vehicle are based upon the information;determining an information-improvement expectation that corresponds to an active-sensi

1. A computer-implemented method comprising: receiving information from one or more sensors of an autonomous vehicle, wherein one or more control processes for the autonomous vehicle are based upon the information;determining an information-improvement expectation that corresponds to an active-sensing action, wherein the active-sensing action comprises an action that is performable by the autonomous vehicle to potentially improve the information upon which at least one of the control processes for the autonomous vehicle is based;determining a risk cost that corresponds to the active-sensing action; andbased on both (i) the information-improvement expectation for the active-sensing action and (ii) the risk cost for the active-sensing action, determining whether the active-sensing action is advisable. 2. The method of claim 1, wherein the at least one sensor comprises one or more of: (a) at least one camera, (b) at least one microphone, (c) a Global Positioning System (GPS), (d) at least one accelerometer, (e) at least one gyroscope, (f) at least one compass, (g) a RADAR, (h) at least one laser rangefinder, (i) LIDAR, (j) at least one steering sensor, (k) a throttle sensor, and (l) at least one brake sensor. 3. The method of claim 1, wherein the active-sensing action comprises one or more of: (a) a movement of the autonomous vehicle from a first lane to a second lane, (b) a change in speed of the autonomous vehicle, (c) a change in position of the autonomous vehicle relative to an aspect of an environment of the autonomous vehicle, (d) a change in position of the autonomous vehicle within a lane, (e) a change in position of at least one of the sensors, (f) a change in operation of at least one of the sensors, and/or (g) a change to the processing of sensor data from at least one of the sensors, wherein the sensor data provides information upon which at least one control process is based. 4. The method of claim 1, wherein determining the information-improvement expectation that corresponds to the active-sensing action comprises: determining an information value of sensor data in a first state of the autonomous vehicle;determining an expected information value of the sensor data that is expected to be provided in a second state of the autonomous vehicle; andsubtracting the information value from the expected information value to determine an information-improvement value for the active sensing action. 5. The method of claim 1, wherein determining the information-improvement expectation that corresponds to the active-sensing action comprises: (i) determining an information-improvement value of a given improvement to the information upon which the at least one of the control process is based;(ii) determining a probability of the given improvement occurring as a result of the autonomous vehicle performing the active-sensing action; and(iii) multiplying the information value of the given improvement by the probability of the given improvement occurring to determine an expected information value for the given improvement. 6. The method of claim 5, wherein the active-sensing action is expected to result in a plurality of improvements to at least one control process, and wherein determining the information-improvement expectation that corresponds to the active-sensing action comprises: performing (i) to (iii) for each of the plurality of improvements to determine an expected information value for each improvement; anddetermining a sum of the expected information values. 7. The method of claim 1, wherein the risk cost is indicative of a risk associated with the autonomous vehicle performing the active-sensing action. 8. The method of claim 1, wherein the risk cost is indicative of a change in risk associated with the autonomous vehicle performing the active-sensing action. 9. The method of claim 1, wherein determining the risk cost that corresponds to the active-sensing action comprises determining a risk penalty for one or more bad events that could occur as a result of the active-sensing action. 10. The method of claim 9, wherein determining the risk penalty for a given bad event comprises: determining a risk magnitude for the given bad event;determining the probability of the given bad event as a result of the active-sensing action; andmultiplying the risk magnitude for the given bad event by the probability of the given bad event occurring to determine the risk penalty for the given bad event. 11. The method of claim 1, wherein determining the risk cost that corresponds to the active-sensing action comprises: determining a risk penalty for each of a plurality of bad events that could occur as a result of the active-sensing action; anddetermining a sum of the risk penalties. 12. The method of claim 1, wherein determining whether the active-sensing action is advisable comprises: determining a score for the active-sensing action based on both (i) the information-improvement expectation and (ii) the risk cost; anddetermining whether the active-sensing action is advisable based on the score. 13. The method of claim 12, wherein determining the score for the active-sensing action comprises subtracting the risk cost from the information-improvement expectation. 14. The method of claim 12, wherein determining whether the active-sensing action is advisable comprises determining whether the score is above a threshold. 15. The method of claim 1, further comprising: determining that the active-sensing action is advisable; andcausing the autonomous vehicle to perform the active-sensing action. 16. The method of claim 1, further comprising, for each of one or more second active-sensing actions, repeating the functions of: determining an information-improvement expectation that corresponds to the second active-sensing action;determining a risk cost that corresponds to the second active-sensing action; andbased on both (i) the information-improvement expectation for the second active-sensing action and (ii) the risk cost for the active-sensing second action, determining whether the second active-sensing action is advisable. 17. The method of claim 1: wherein the active-sensing action comprises a movement of the autonomous vehicle into a different lane of a road to potentially improve a representation of a traffic light in image data from a camera; andwherein determining the information-improvement expectation that corresponds to the active-sensing action comprises determining an expected improvement to a process that is based on a traffic-light state, wherein the expected improvement corresponds to the movement of the autonomous vehicle into the different lane of the road. 18. The method of claim 1: wherein the active-sensing action comprises a movement of the autonomous vehicle into a different position relative to at least one other vehicle to potentially improve environmental information provided by image data from a camera; andwherein determining the information-improvement expectation comprises determining an expected improvement to a process that is based on the environmental information, wherein the expected improvement corresponds to the movement of the autonomous vehicle into the different position relative to the at least one other vehicle. 19. The method of claim 1: wherein the active-sensing action comprises an action by the autonomous vehicle to potentially improve a view of a GPS satellite; andwherein determining the information-improvement expectation that corresponds to the active-sensing action comprises determining an expected improvement to a process that is based on data from a GPS signal, wherein the expected improvement corresponds to the action by the autonomous vehicle to improve the view of the GPS satellite. 20. An autonomous-vehicle system comprising: one or more sensors;a computer system configured to: use information provided by the one or more sensors for one or more control processes of an autonomous vehicle;determine an information-improvement expectation that corresponds to an active-sensing action, wherein the active-sensing action comprises an action that is performable by the autonomous vehicle to potentially improve the information upon which at least one of the control processes for the autonomous vehicle is based;determine a risk cost that corresponds to the active-sensing action; andbased on both (i) the information-improvement expectation for the active-sensing action and (ii) the risk cost for the active-sensing action, determine whether the active-sensing action is advisable. 21. The system of claim 20, wherein, to determine the information-improvement expectation that corresponds to the active-sensing action, the computer system is configured to: determine an information value of sensor data in a first state of the autonomous vehicle;determine an expected information value of the sensor data that is expected to be provided in a second state of the autonomous vehicle; andsubtract the information value from the expected information value to determine an information-improvement value for the active sensing action. 22. The system of claim 20, wherein, to determine the information-improvement expectation that corresponds to the active-sensing action, the computer system is configured to: determine an information-improvement value of a given improvement to the information upon which the at least one of the control process is based;determine a probability of the given improvement occurring as a result of the autonomous vehicle performing the active-sensing action; andmultiply the information value of the given improvement by the probability of the given improvement occurring to determine an expected information value for the given improvement. 23. The system of claim 20, wherein, to determine the risk cost that corresponds to the active-sensing action, the computer system is configured to determining a risk penalty for one or more bad events that could occur as a result of the active-sensing action. 24. The system of claim 23, wherein, to determine the risk penalty for a given bad event, the computer system is configured to: determine a risk magnitude for the given bad event;determine the probability of the given bad event as a result of the active-sensing action; andmultiply the risk magnitude for the given bad event by the probability of the given bad event occurring to determine the risk penalty for the given bad event. 25. The system of claim 20, wherein, to determine whether the active-sensing action is advisable, the computer system is configured to: determine a score for the active-sensing action based on both (i) the information-improvement expectation and (ii) the risk cost; anddetermine whether the active-sensing action is advisable based on the score. 26. A non-transitory computer-readable medium having program instructions stored thereon that are executable by at least one processor, the program instructions comprising: instructions for receiving information from one or more sensors of an autonomous vehicle, wherein one or more control processes for the autonomous vehicle are based upon the information;instructions for determining an information-improvement expectation that corresponds to an active-sensing action, wherein the active-sensing action comprises an action that is performable by the autonomous vehicle to potentially improve the information upon which at least one of the control processes for the autonomous vehicle is based;instructions for determining a risk cost that corresponds to the active-sensing action; andinstructions for, based on both (i) the information-improvement expectation for the active-sensing action and (ii) the risk cost for the active-sensing action, determining whether the active-sensing action is advisable. 27. The non-transitory computer-readable medium of claim 26, wherein the instructions for determining the information-improvement expectation that corresponds to the active-sensing action comprise: instructions for determining an information value of sensor data in a first state of the autonomous vehicle;instructions for determining an expected information value of the sensor data that is expected to be provided in a second state of the autonomous vehicle; andinstructions for subtracting the information value from the expected information value to determine an information-improvement value for the active sensing action. 28. The non-transitory computer-readable medium of claim 26, wherein the instructions for determining the information-improvement expectation that corresponds to the active-sensing action comprise: (i) instructions for determining an information-improvement value of a given improvement to the information upon which the at least one of the control process is based;(ii) instructions for determining a probability of the given improvement occurring as a result of the autonomous vehicle performing the active-sensing action; and(iii) instructions for multiplying the information value of the given improvement by the probability of the given improvement occurring to determine an expected information value for the given improvement. 29. The non-transitory computer-readable medium of claim 26, wherein the instructions for determining the risk cost that corresponds to the active-sensing action comprise instructions for determining a risk penalty for one or more bad events that could occur as a result of the active-sensing action. 30. The non-transitory computer-readable medium of claim 29, wherein the instructions for determining the risk penalty for a given bad event comprise: instructions for determining a risk magnitude for the given bad event;instructions for determining the probability of the given bad event as a result of the active-sensing action; andinstructions for multiplying the risk magnitude for the given bad event by the probability of the given bad event occurring to determine the risk penalty for the given bad event. 31. The non-transitory computer-readable medium of claim 26, wherein the instructions for determining whether the active-sensing action is advisable comprise: instructions for determining a score for the active-sensing action based on both (i) the information-improvement expectation and (ii) the risk cost; andinstructions for determining whether the active-sensing action is advisable based on the score.