Tuning a safety system based on near-miss events
An autonomous vehicle safety system may activate to prevent collisions by detecting that a planned trajectory may result in a collision. If the safety system is overly sensitive, it may cause false positive activations, and if the system isn't sensitive enough the collision avoidance system may not activate and prevent a collision, which is unacceptable. It may be impossible or prohibitively difficult to detect false positive activations of a safety system and it is unacceptable to risk a false negative, so tuning the safety system is notoriously difficult. Tuning the safety system may include detecting near-miss events using surrogate metrics, and tuning the safety system to increase or decrease a rate of near-miss events as a stand-in for false positives.
1. A method comprising:
receiving sensor data associated with operation of a safety component of an autonomous vehicle;
determining, based at least in part on the sensor data, a representative metric associated with operation of the autonomous vehicle;
determining, based at least in part on the representative metric, that at least a portion of the operation is associated with a near-miss event, wherein the near-miss event indicates that the autonomous vehicle passed within a threshold distance of an object;
altering a parameter of one or more of the safety component or another component of the autonomous vehicle based at least in part on the near-miss event to determine an updated component, wherein the altering:
is based at least in part on a near-miss event count and further comprises determining that the safety component remained inactive; and
decreases a near-miss rate below an upper near-miss threshold and increases a false positive rate, or
increases the near-miss rate to meet or exceed a lower near-miss threshold and decreases the false positive rate; and
transmitting the altered parameter to the autonomous vehicle, such that the autonomous vehicle controls operations of the autonomous vehicle based at least in part on the altered parameter.
2. The method of claim 1 , further comprising:
determining the upper near-miss threshold based at least in part on a first near-miss count or false positive count per miles driven; and
determining the lower near-miss threshold based at least in part on a second near-miss count per miles driven, wherein the second near-miss count per miles driven is less than the first near-miss count per miles driven.
3. The method of claim 1 , wherein the representative metric comprises one or more of:
an encroachment time,
a gap time,
a deceleration rate to avoid a collision with an object,
a stopping distance required before colliding with an object,
a proportion between a remaining distance to a predicted point of collision to a minimum stopping distance,
a time to collision,
a lateral distance to a closest object,
a longitudinal acceleration, or
a longitudinal and/or lateral jerk.
4. The method of claim 1 , wherein the altering comprises altering one or more of:
a minimum distance between the autonomous vehicle and a detected object;
a period of time in which the autonomous vehicle is predicted to reach a position of the detected object;
a maximum acceleration associated with operation of the vehicle; or
a prediction or heuristic of an object's ability to stop or modify the object's behavior.
5. The method of claim 1 further comprising:
determining, based at least in part on the sensor data, an additional representative metric associated with the operation of the autonomous vehicle;
determining, based at least in part on the additional representative metric, that at least a second portion of the operation is associated with normative operation;
determining that the safety component activated during the second portion of the operation; and
altering a second parameter of the one or more of the safety component or another component of the autonomous vehicle to determine a second updated component based at least in part on determining that the second portion is associated with normative operation and that the safety component activated.
6. The method of claim 5 , wherein, upon simulating or operating the autonomous vehicle in a same or similar scenario to a scenario associated with the second portion of the operation, altering the second parameter causes the safety component to abstain from activating, resulting in a second near-miss event.
7. A system comprising:
one or more processors; and
a memory storing processor-executable instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
receiving sensor data associated with operation of a safety component of an autonomous vehicle;
determining, based at least in part on the sensor data, one or more characteristics of operation of the autonomous vehicle;
determining, based at least in part on the one or more characteristics, that at least a portion of the operation is associated with a near-miss event;
altering a parameter of the one or more of the safety component or another component of the autonomous vehicle based at least in part on the near-miss event to determine an updated component, wherein altering the parameter is based at least in part on a near-miss event count and further comprises determining that the safety component remained inactive; and
transmitting the altered parameter to the autonomous vehicle, such that the autonomous vehicle controls operations of the autonomous vehicle based at least in part on the altered parameter.
8. The system of claim 7 , wherein the altering:
decreases a near-miss rate below an upper near-miss threshold and increases a false positive rate, or
increases the near-miss rate to meet or exceed a lower near-miss threshold and decreases the false positive rate.
9. The system of claim 8 , wherein the operations further comprise:
determining the upper near-miss threshold based at least in part on a first near-miss count or false positive count per miles driven; and
determining the lower near-miss threshold based at least in part on a second near-miss count per miles driven, wherein the second near-miss count per miles driven is less than the first near-miss count per miles driven.
10. The system of claim 7 , wherein the altering comprises altering one or more of:
a minimum distance between the autonomous vehicle and a detected object;
a period of time in which the autonomous vehicle is predicted to reach a position of the detected object;
a maximum acceleration associated with operation of the vehicle; or
a prediction or heuristic of an object's ability to stop or modify the object's behavior.
11. The system of claim 7 , wherein the operations further comprise:
determining, based at least in part on the sensor data, one or more second characteristics associated with the operation of the autonomous vehicle;
determining, based at least in part on the one or more second characteristics, that at least a second portion of the operation is associated with normative operation;
determining that the safety component activated during the second portion of the operation; and
altering a second parameter of the one or more of the safety component or another component of the autonomous vehicle to determine a second updated component based at least in part on determining that the second portion is associated with normative operation and that the safety component activated.
12. The system of claim 11 , wherein, upon simulating or operating the autonomous vehicle in a same or similar scenario to a scenario associated with the second portion of the operation, altering the second parameter causes the safety component to abstain from activating, resulting in a second near-miss event.
13. A non-transitory computer-readable medium storing processor-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving sensor data associated with operation of a safety component of an autonomous vehicle;
determining, based at least in part on the sensor data, one or more characteristics of operation of the autonomous vehicle;
determining, based at least in part on the one or more characteristics, that at least a portion of the operation is associated with a near-miss event;
altering a parameter of the one or more of the safety component or another component of the autonomous vehicle based at least in part on the near-miss event to determine an updated component, wherein altering the parameter is further based at least in part on a near-miss event count and further comprises determining that the safety component withheld activation; and
transmitting the altered parameter to the autonomous vehicle, such that the autonomous vehicle controls operations of the autonomous vehicle based at least in part on the altered parameter.
14. The non-transitory computer-readable medium of claim 13 , wherein the altering:
decreases a near-miss rate below an upper near-miss threshold and increases a false positive rate, or increases the near-miss rate to meet or exceed a lower near-miss threshold target and decreases the false positive rate.
15. The non-transitory computer-readable medium of claim 14 , wherein the operations further comprise:
determining the upper near-miss threshold based at least in part on a first near-miss count or false positive count per miles driven; and
determining the lower near-miss threshold based at least in part on a second near-miss count per miles driven, wherein the second near-miss count per miles driven is less than the first near-miss count per miles driven.
16. The non-transitory computer-readable medium of claim 13 , the altering comprises altering one or more of:
a minimum distance between the autonomous vehicle and a detected object;
a period of time in which the autonomous vehicle is predicted to reach a position of the detected object;
a maximum acceleration associated with operation of the vehicle; or
a prediction or heuristic of an object's ability to stop or modify the object's behavior.
17. The non-transitory computer-readable medium of claim 13 , wherein the operations further comprise:
determining, based at least in part on the sensor data, one or more second characteristics associated with the operation of the autonomous vehicle;
determining, based at least in part on the one or more second characteristics, that at least a second portion of the operation is associated with normative operation;
determining that the safety component activated during the second portion of the operation; and
altering a second parameter of the one or more of the safety component or another component of the autonomous vehicle to determine a second updated component based at least in part on determining that the second portion is associated with normative operation and that the safety component activated.
18. The non-transitory computer-readable medium of claim 17 , wherein, upon simulating or operating the autonomous vehicle in a same or similar scenario to a scenario associated with the second portion of the operation, altering the second parameter causes the safety component to abstain from activating, resulting in a second near-miss event.