U.S. patent application number 15/619581 was filed with the patent office on 2017-09-28 for personalize self-driving cars.
The applicant listed for this patent is Xiaoning Huai. Invention is credited to Xiaoning Huai.
Application Number | 20170274908 15/619581 |
Document ID | / |
Family ID | 59897697 |
Filed Date | 2017-09-28 |
United States Patent
Application |
20170274908 |
Kind Code |
A1 |
Huai; Xiaoning |
September 28, 2017 |
Personalize self-driving cars
Abstract
A method to personalize the operation of a self-driving
automobile is disclosed that improves the applicability and user
appreciation of a self-driving automobile by acquiring and applying
the user preference data set and the user profile data set,
incorporating individual user choice of preferred driving behaviors
on different scenarios and user preferred driving styles, and/or
the moral or ethics into the control of the automobile
operation.
Inventors: |
Huai; Xiaoning; (Sunnyvale,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Huai; Xiaoning |
Sunnyvale |
CA |
US |
|
|
Family ID: |
59897697 |
Appl. No.: |
15/619581 |
Filed: |
June 12, 2017 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
H04L 67/306 20130101;
G05D 1/0221 20130101; B60W 50/0098 20130101; B60W 2050/0089
20130101; B60W 50/08 20130101; G05D 1/0214 20130101; B60W 2050/0082
20130101; B60W 2540/043 20200201; G06K 9/6262 20130101; B60W 40/09
20130101; G05D 1/0088 20130101 |
International
Class: |
B60W 40/09 20060101
B60W040/09; H04L 29/08 20060101 H04L029/08; G05D 1/00 20060101
G05D001/00 |
Claims
1. A method of personalizing a self-driving car comprising a robot
of the car conducting steps of identifying a user; acquiring the
user preference data set and/or the user profile data set;
acquiring the preferred driving styles, and/or the moral or ethics
traits of the user; and applying the user preference data set
and/or the preferred driving style and/or the moral or ethics
traits of the user in operating the car.
2. The method of claim 1, wherein acquiring the user preference
data set comprising the robot identifying a user; presenting a
collection of roadway and traffic conditions scenarios one by one,
inviting the user inputting his or her choices on a preferred
handling behavior, and inputting the scenario/choice pairs in an
entry of the user preference data set through an interactive
initialization process between the robot and the user at the time
of purchasing or requesting the service of the car; or by receiving
a previously acquired user preference data set of the user and
confirming or updating the acquired data prior to or at the time of
the self-driving being used in a public roadway.
3. The method of claim 1, wherein acquiring the user profile data
set comprising the robot identifying a user; acquiring personal
information provided by the user through an interactive interface
between the robot and the user and/or by researching the public
records through a wireless communication system or an electronic
media device, or through receiving the user profile data set
through a wireless communication system or an electronic media
device.
4. The method of claim 1, wherein acquiring the preferred driving
styles, and/or the moral or ethics traits of the user comprising
extracting the preferred driving styles, and/or the moral or ethics
traits of the user from the acquired user preference data set
and/or the user profile data set or receiving the extracted
preferred driving styles, and/or the moral or ethics traits of the
user.
5. The method of claim 2, wherein the collection of roadway and
traffic conditions scenarios are categorized into segments
comprising the blinking zone, the emergency zone and the cruise
zone based on the estimated response time of the car to roadway or
traffic scenarios.
6. The method of claim 1, wherein applying the user preference data
set and/or the preferred driving style and/or the moral or ethics
traits of the user in operating the self-driving car comprising
restraining the operating to be lawful; finding a closest match
between a current scenario and a scenario used in the collection
for acquisition of the user preference data set, and applying the
user preference data operating the car if a match being found close
enough; generating a suggestion how to handle a scenario based on
the user preferred driving style and/or the moral or ethics traits
if a close enough match not found, and achieving an optimal
solution by considering the suggestion together with other options
generated by AI control subsystem.
7. The method of claim 1, wherein the user preference data set
and/or the preferred driving style and/or the moral or ethics
traits comprising the data sets of one of the users riding the car,
or of a designated user not riding the car or the factory
settings.
8. (canceled)
9. The method of claim 2, wherein the user inputting his or her
choice on a preferred handling behavior on a roadway and traffic
conditions scenario comprising selecting an answer among multiple
choices or answering to a yes or no question or entering a numeric
value within a normalized range, indicating a percentage degree of
a consent or discontent to a answer.
10. The method of claim 1, wherein the personalizing continues
during the driving, comprising the robot executing guidance from a
user in operation of the car and updating the user preference data
set by the roadway and traffic scenario/guidance data pairs,
through interactions between the robot and the user over roadway
and traffic scenarios.
11. The method of claim 1, wherein the personalizing continues
during the driving, comprising the robot automatically detecting
and analyzing the facial and/or body languages of a user reflecting
his or her sentiment to the behaviors of the car; tuning the
operations of the car; extracting the user profile data and
updating the user profile data set.
Description
TECHNICAL FIELD
[0001] Artificial intelligence, self-driving cars, and robot.
BACKGROUND
[0002] AI (denotes artificial intelligence hereby and hereafter in
this disclosure) based driving automation has evolved now to a
stage of heavy premarketing road test by several self-driving car
manufacturers. Among other issues, accidents are still occasionally
reported calling for more improvements. A self-driving car could be
viewed as if a robot sits on a conventional car, though it does not
take the shape of what is commonly presented or perceived,
comprising a sensing sub-system, an AI control sub-system and an
activation sub-system, and the conventional car should be altered
significantly for a better integration, as illustrated in FIG. 1. A
self-driving car drives itself from one start point to a
destination set by a user or a remote controller through a wireless
communication system or an electronic media device and guided by a
GPS navigational system with or without involving a user in the
car. It can contain one or more passengers or no passengers, for
example when it is sent for a passenger. The robot conducts
real-time scene analysis of roadway and traffic events, interprets
and applies the traffic rules wherever it is driving, and
synchronizes activation subsystem to make the driving for a user
(denotes an owner or a passenger who rides or uses the self-driving
car hereby and hereafter in this disclosure). A robot on the
self-driving car has been trained in the factory and learned the
general skills and rules of a car operation. However, driving as a
human activity has many quality attributes than just moving or
transportation, such as safety, comfort, exercise, sport and so on,
which are valued according to each user's experiences, favors,
moral and ethics traits among other things. A personalized
self-driving car could do better to satisfy a user's needs, and a
method for its customer design is hereby introduced in this
disclosure.
SUMMARY OF THE INVENTION
[0003] A method is disclosed to personize a self-driving car's
driving behavior to reflect user's preferred driving styles, and/or
moral or ethics traits in handling normal traffic and emergency
scenarios, based on acquiring and analyzing user preference data
and user profile data to start with and a continuing leaning by the
robot during the driving.
BRIEF DISCUSSION OF DRAWINGS
[0004] FIG. 1 Illustration of a functional structure of a
self-driving car.
[0005] FIG. 2 Illustration of categorized response time interval to
roadway and traffic events, the shaded area around T1 and T2
indicate it should be considered as a zone with a boundary varying
from model to model, and from time to time.
[0006] FIG. 3 Illustration of personalizing a self-driving car
procedures.
[0007] FIG. 4 Table 1, an example of impact on operation by
personalized user preference data.
[0008] FIG. 5 Illustration of how to apply user data on roadway and
traffic scenarios.
DETAILED DESCRIPTION OF THE INVENTION
[0009] The robot of a self-driving car keeps monitoring and
detecting roadway and traffic conditions by its sensing sub-system,
and any events prompting for a responding adjustment of its driving
will be analyzed to fall into one of the three categorized response
time intervals, taking into account the distance of an involved
object to and the speed of the car, the time needed for the robot
to run algorithms and activation sub-system, and for the activation
to take effect, as illustrated in FIG. 2, wherein the actual
parameter values could be different from one car model to another,
since each model usually has a performance features of
maneuverability as designed by its manufacturer. The interval
between time 0 to T1 is referred to as The Blinking Zone, wherein
the robot can virtually do little or nothing to address the event
or avoid an accident but could act somehow to minimize the damages
to the user or the car and send out alarms if there is an accident.
The interval between T1 to T2 is referred to as The Emergency Zone,
wherein actions could be taken to address the events or avoid an
accident or let an accident happen in one way or another that would
put different risks of damages to the driver, the car of the driver
and/or a third party who is involved in the accident, such as a
vehicle or a pedestrian who happens to share the roadway. The
interval from T2 beyond is referred to as the Cruise Zone, wherein
the roadway and traffic events are easily manageable and chance of
an accident is very small. Corresponding to each interval, there
are pre-acquired personalized sets of data for each user reflecting
the user's choices of behaviors in different scenarios, preferred
driving styles, and/or moral or ethics traits, which will be used
by the robot in its control of the operations, a process hereby
referred to as a personalized self-driving and is detailed
below.
[0010] Personalize self-driving starts by an initialization
process, which takes place before the car is started or moved for a
roadway driving, using an interactive interface to communicate
between the user and the robot of a self-driving car as illustrated
in FIG. 3. A user should be identified first for the robot to
acquire user preferences data set of preferred behaviors of a
self-driving car on a collection of roadway and traffic scenarios.
There are many ways to identify a user by state of the art
technologies, and a user ID/password combination could be an easy
one. The robot would present a collection of roadway and traffic
conditions scenarios one by one, and invite the user inputting his
or her opinions on a preferred handling behavior by selecting an
answer among multiple choices or answering to a yes or no question.
Since it is very difficult and lengthy to cover all possible
scenarios, some generalization and categorization of scenarios are
necessary, and a numeric value within a normalized range of
indicating a percentage degree of a consent or discontent to a
choice of answer is optionally used. The interactive interface
between the robot and the user could be of a visual media such as a
touch screen panel for display and input, or an audio media such as
a speaker announcement combined with a microphone and a speech
recognition module to take the inputs, or a combination thereof,
for users without vision or hearing disabilities. For user with
disabilities, however, an assistant to the user could help with the
initialization to use the above common interface mechanisms for the
communication, or an adaptive device could be designed and
installed. The acquired preference data will then be stored in a
data structure named the user preference data set, which has an
entry for each user of the self-driving car categorized according
to the above described three segments of response time
intervals.
[0011] In addition to a user preference data set, a second data set
named the user profile data set is also acquired, based on
information provided by a user and/or through a research by the
robot through a wireless communication system or an electronic
media device, which comprises the age, gender, profession,
education level and other personal and/or public information
available such as marriage status, living areas, driving, credit,
insurance, health and criminal records. The acquisition of user
profile data set could take place between the robot and a user
using an interactive interface at the time of purchasing or
requesting a service of a self-driving car. After a user provides
related information, the robot runs a background check using a
wireless communication system or an electronic media device.
Alternatively, these data could be acquired prior to purchasing or
using the service of a self-driving car between a user and a vender
or service provider and delivered to the robot of a self-driving
car later.
[0012] The robot will then analyze these two sets of user data to
determine and profile preferred driving style, and/or moral or
ethics traits of a user and infer the proper behavior for the
self-driving car in a variety of difficult roadway and traffic
scenarios based on data from behavior modeling, factory tests and
user statistics and related algorithms, and store the results in a
data structure in the user profile data set. Alternatively, the two
sets of data could also be analyzed using resources elsewhere and
the results are delivered to the robot later. An illustration how
to apply these data for real time operation of a self-driving car
in a user's personal way can be found in FIG. 5. In the first
place, a self-driving should follow the traffic rules and other
laws regarding a vehicle operation. Within that restraint, one
recommendation is to run scenario matching first, and find the
closest match between the current scenario and a scenario used in
the collection for acquisition of the user preference data, and
apply the user preference data operating the car if the match is
close enough. However, when a close enough match could not be
found, the robot should then refer to the user preferred driving
style and/or the moral or ethics traits to generate a suggestion
how to handle an unexpected scenario and find an optimal solution
by considering it together with other options generated by AI
control subsystem. So, there is a clear difference between how to
use these two sets of user data. Certain restrictions are applied
as a default settings for the self-driving cars in general. For
example, since this disclosure is not concerned about the
application to use the driver-less technology as a battle vehicle
in a war or as a vehicle for law enforcement, the self-driving car
is recommended to be inhibited to be engaged in any offensive
action against any third parties, including pedestrians, other
vehicles etc. It should also be barred from any self-destruction
behavior such as running out of a cliff or against a road barrier
or walls of a building, unless the AI control sub-system of the
robot determines such a move is necessary for reducing the
seriousness of an otherwise unavoidable accident and the user has
optioned such a choice in the user preference data set. Although in
general, applying the user preference data and user's preferred
driving styles, and/or moral or ethics traits is intended to
satisfy the user's expectation, there are exceptions on the
contrary, for example, if a user riding the car is found to be
drunk by an alcoholic sensor, or to be a habitual reckless driving
offender, certain functions such as user overriding the robot for
operating the car should be restricted.
[0013] when multiple users are riding the car, it is optional to
select the user preference data and the user profile data of one of
the riders in assisting the operation of the car. In case there is
no passenger riding the car, a self-driving car will follow its
factory settings or use a pre-acquired designated user preference
data set and user profile data set.
[0014] An example of impact on operation by personalized user
preference data is illustrated in FIG. 4 Table 1. Since how to
handle emergency between T1 and T2 is most critical and
controversial to the safety behavior of a self-driving car, some
examples are designed and given below as an illustration of
scenarios and preference data pairs.
EXAMPLE 1
[0015] A self-driving car is driving on a roadway at a normal speed
approaching an intersection with a green light, a bicycle suddenly
runs red light from one side of the roadway appearing in front of
the self-driving car. The robot finds braking the car is too late
to avoid the accident, but the car to the left or right might have
a chance, which would violate the traffic rules by running into a
wrong lane and have a chance to damage the self-driving car, what
would be the user's opinion? The choices for the answer are: [0016]
A. Brake the car [0017] B. Swing the car
EXAMPLE 2
[0018] When a self-driving car entering a potential accident
involving another party that might have the liability for causing
the accident, to what degree of risk between 0 and 1 would you take
to avoid the accident, if the self-driving car has been following
the traffic rules?
EXAMPLE 3
[0019] When a collision between the self-driving car and another
vehicle is not avoidable, which of the following you would choose?
[0020] A. Minimize the damage to yourself no matter what happens to
the other party [0021] B. Minimize the damage to yourself no matter
what happens to the other party if the other party has the
liability [0022] C. Take some risk of damaging yourself depending
the circumstances to reduce the damage to the other party
EXAMPLE 4
[0023] When an accident is not avoidable, which of the following
you would choose? [0024] A. Minimize the damage to the passenger
sitting on the front-left seat [0025] B. Minimize the damage to the
passenger sitting on the back-right seat [0026] C. Minimize the
damage to myself no matter where I am sitting
EXAMPLE 5
[0027] Your preferred driving style in highway is: [0028] A. Quick
and fast [0029] B. Steady and smooth
[0030] A continuing user adaptation by learning during the driving
is illustrated in FIG. 3 by module 380, particularly if the user is
a recurrent one such as an owner of the car, and data acquired in
the initialization may not cover all roadway and traffic scenarios
and the interpretation by the robot of the user preference data may
not fully satisfy the user. The robot could prompt messages or make
announcements, through visual or sound or other kind of media
devices, about an unfamiliar or untrained or hazardous roadway and
traffic condition, and asks for the user to input a guidance or
command, and executes the operation accordingly upon receiving such
a guidance or command, updating the user preference data set by
taking the user's inputs in combination with the scenarios as a
result. The user could also take over the driving physically when
necessary and if it is feasible in the design, or take initiatives
to direct or correct the driving behavior of the robot through an
interactive interface. On the other hand, the robot could use a
gaze, gesture or other monitoring techniques to detect and analyze
the user's body languages reflecting his or her experiences during
the driving, and tune its operation accordingly. Thereby, a
satisfactory personalized match between the behavior of the robot
of a self-driving car and the expectation of its user could be
realized.
* * * * *