SYSTEMS AND METHODS FOR HAZARD MITIGATION
A system and method to avoid collisions on highways, and to minimize the fatalities, injury, and damage when a collision is unavoidable. The system includes sensor means to detect other vehicles, and computing means to evaluate when a collision is imminent and to determine whether the collision is avoidable. If the collision is avoidable by a sequence of controlled accelerations and decelerations and steering, the system implements that sequence of actions automatically. If the collision is unavoidable, a different sequence is implemented to minimize the overall harm of the unavoidable collision. The system further includes indirect mitigation steps such as flashing the brake lights automatically. An optional post-collision strategy is implemented to prevent secondary collisions, particularly if the driver is incapacitated. Adjustment means enable the driver to set the type and timing of automatic interventions.
1 . A non-transitory computer-readable medium comprising instructions for causing a computing environment mounted on a subject vehicle to perform a method for mitigating an imminent collision, comprising:
receiving a signal from a sensor mounted on the subject vehicle, the signal indicating presence and trajectory of a second vehicle;
using a processor, causing a throttle, brakes, or steering, or a combination of these, to change the motion of the subject vehicle according to a first sequence of actions to avoid a collision with the second vehicle, the first sequence of actions based at least in part on the received signal;
based at least in part on the first sequence of actions and subsequent signals from the sensor, causing a throttle, brakes, or steering, or a combination of these, to change the motion of the subject vehicle according to a second sequence of actions to minimize the harm of a collision if the collision is determined to be unavoidable, wherein the determination as to whether a collision is avoidable is performed by a processor on the subject vehicle or by a cloud server in signal communication with the subject vehicle, wherein the first or second sequences of actions are further determined at least in part based on machine learning or artificial intelligence, the machine learning or artificial intelligence based on prior learning, the machine learning or artificial intelligence used to determine additional data by a step selected from the group consisting of: inferring an intent of a driver of the subject or second vehicle or both, predicting a future position or velocity of the subject or second vehicle, determining whether a collision is avoidable, determining or identifying driving habits of the driver of the subject vehicle or second vehicle.
2 . The medium of claim 1 , wherein the machine learning algorithm is configured to learn based on observation.
3 . The medium of claim 2 , wherein the machine learning algorithm is configured to adjust model parameters based on observed instances, whereby improved predictions may be made for future scenarios.
4 . The medium of claim 1 , wherein the machine learning algorithm is configured to learn based on experience.
5 . The medium of claim 4 , wherein the machine learning algorithm is configured to adjust model parameters based on experience, whereby optimal safety may be accorded in future scenarios.
6 . The medium of claim 5 , whereby the experience includes monitoring a driver of the subject vehicle and adjusting model parameters based on a set of determined habits of the driver.
7 . A non-transitory computer-readable medium in a subject vehicle comprising instructions for causing a computing environment to perform a method for mitigating vehicle collisions, comprising:
processing data from one or more external sensors, and from the processing detecting a second vehicle proximate to the subject vehicle;
calculating future trajectories of the subject and second vehicle, and from that calculating determining whether the subject and second vehicle will collide if no corrective action is taken;
calculating whether the collision can be avoided if the subject vehicle is caused to change trajectory according to one or more first sequences of actions, and if so implementing a sequence that avoids the collision; and
if none of the sequences of actions is determined to avoid the collision, then implementing a second sequence of actions configured to minimize the harm of the collision;
wherein each sequence of actions comprises at least one or more separate sequential actions; and
wherein each action comprises a specified acceleration or deceleration or steering action with a specified intensity, or a waiting period, occurring at a specified time and for a specified duration, and
wherein the first or second sequences of actions are further determined at least in part based on machine learning or artificial intelligence, the machine learning or artificial intelligence based on prior learning, the machine learning or artificial intelligence used to determine additional data by a step selected from the group consisting of: inferring an intent of a driver of the subject or second vehicle or both, predicting a future position or velocity of the subject or second vehicle, determining whether a collision is avoidable, determining or identifying driving habits of the driver of the subject vehicle or second vehicle.
8 . The method of claim 7 , further comprising:
reading a plurality of previously-prepared sequences of actions from non-transient computer-readable media; and
selecting a one or more of the previously-prepared sequences, the selection based on the processed data.
9 . The method of claim 8 , wherein the selected sequences comprise sequences of actions that have successfully mitigated collisions that have processed data corresponding to the currently projected collision.
10 . The method of claim 7 , wherein the actions include at least two of:
periods of positive acceleration;
periods of braking;
periods of steering; and
periods of waiting.
11 . The method of claim 7 , wherein the actions include conditional instructions.
12 . The method of claim 11 , wherein the conditional instructions include at least one of:
changing the speed of the subject vehicle to match the speed of another vehicle;
changing the speed of the subject vehicle to match the average of two other vehicle speeds;
changing the position of the subject vehicle relative to another vehicle; and
changing the speed of the subject vehicle responsive to the brake lights of another vehicle.
13 . The method of claim 7 , further comprising instructing the subject vehicle driver to take an action indicating responsiveness after a collision.
14 . The method of claim 13 , wherein the action indicating responsiveness is selected from the group consisting of: pressing the brake pedal, pressing the accelerator pedal, turning the steering wheel, and pressing a button.
15 . The method of claim 7 , further comprising:
determining whether the subject vehicle driver is still responsive after a collision; and
driving the subject vehicle to a side of a road if the subject vehicle driver is determined to be unresponsive.
16 . The method of claim 7 , further comprising:
determining that the subject vehicle driver is unresponsive;
while the subject vehicle driver is unresponsive, autonomously driving the subject vehicle toward a side of a road; and
detecting that the subject vehicle driver has performed an action indicating responsiveness; and
relinquishing control of the subject vehicle to the subject vehicle driver.
17 . A non-transitory computer readable medium, comprising instructions for causing a computing environment to perform a method for mitigating vehicle collisions, the method comprising:
detecting, with a sensor mounted on a subject vehicle, a second vehicle, and calculating if the second vehicle will collide with the subject vehicle;
calculating whether a first set of actions would avoid a collision with the subject vehicle;
if the first set of actions would avoid the collision, implementing the first set of actions;
if the first set of actions would fail to avoid the collision, calculating whether a second set of actions would avoid the collision:
if the second set of actions would avoid the collision, implementing the second set of actions; and
if the second set of actions would fail to avoid the collision, implementing which of the first or second set of actions, or a third set of actions, that causes a least amount of harm;
wherein each set of actions comprises a set of sequential accelerations of the subject vehicle, each acceleration having at least a specified timing and duration;
and wherein a collision comprises the second vehicle striking the subject vehicle or passing within a predetermined radius of the subject vehicle, and
wherein the first or second sequences of actions are further determined at least in part based on machine learning or artificial intelligence, the machine learning or artificial intelligence based on prior learning, the machine learning or artificial intelligence used to determine additional data by a step selected from the group consisting of: inferring an intent of a driver of the subject or second vehicle or both, predicting a future position or velocity of the subject or second vehicle, determining whether a collision is avoidable, determining or identifying driving habits of the driver of the subject vehicle or second vehicle.
18 . The medium of claim 17 , wherein the method is such that the least harm is a local minimum or an absolute minimum of harm.
19 . The medium of claim 17 , wherein the method is such that the least amount of harm is a smallest amount of harm among sets of actions yet calculated.
20 . The medium of claim 17 , wherein the method is such that the least amount of harm is a lowest amount of harm achievable by accelerating, decelerating, and/or steering the subject vehicle.