Integrated Safety - Flash Memory Summit

Integrated Safety - Flash Memory Summit

2015.8.12_Flash Memory_Summit_2015, Santa Clara
Driving Intelligence for Safer Automobiles
Hideo Inoue
Toyota Motor Corporation
Integrated Safety
1/30
The History of the Automobile
1769
The world's first automobile:
A steam-powered vehicle
Today (from 2006)
Pre-collision system (PCS)
Emergency avoidance of
objects and pedestrians
Invented by Nicolas-Joseph Cugnot
Integrated Safety
2/30
Our perspectives of safe & secure driving
Tomorrow
Driving quality enhancement based on safety
technologies
Today
?
(2) Emergency avoidance
(1) Mitigation
*Rear-end collision avoidance
*Pedestrian protection
*Head on collisions
*PCS, ESC, ABS, etc.
*Airbag
*Seatbelt
*Body structure
Time
Integrated Safety
3/30
Our perspectives of safe & secure driving
Tomorrow
Driving quality enhancement based on safety
technologies
Today
(3) Avoidance of human-error
induced accidents
1) Distraction, fatigue, drowsiness, heart
attacks, etc.
2) Less stress in monotonous driving
situations
e.g., highway/interstate driving/trips,
congested/stop-and-go traffic
3) Elderly/disabled drivers
(2) Emergency avoidance
(1) Mitigation
*Rear-end collision avoidance
*Pedestrian protection
*Head on collisions
*PCS, ESC, ABS, etc.
*Airbag
*Seatbelt
*Body structure
Time
Integrated Safety
4/30
Our research into intelligent driving technology
 Monitor around the vehicle and predict risks.
 Find and follow safe routes
From 2005:
From 2007:
Learn from skilled drivers
Pursue sensing, perception,
 Verify potential of autonomous
and recognition technology
vehicle dynamic controls
Integrated Safety
5/30
Two topics for tomorrow
1. ADAS with Driving Intelligence for a Safer and
More Secure Traffic Society for Elderly Drivers
(S-Innovation Project in Japan)
2. Traffic Congestion and Highway Driving
Intelligence
Integrated Safety
6/30
Two topics for tomorrow
1. ADAS with Driving Intelligence for a Safer and
More Secure Traffic Society for Elderly Drivers
(S-Innovation Project in Japan)
2. Traffic Congestion and Highway Driving
Intelligence
Integrated Safety
7/30
Motivation and objectives: Overcoming fear of driving
- Deterioration in driving ability reduces self-confidence in driving.
- However, elderly drivers are highly motivated to improve QOL.
- Intelligent driving systems can help to compensate for this deterioration in
physical ability and overcome the fear of driving.
Recognized the warning and operated brake.
Recognized the warning but did not brake.
Did not recognize the warning and did not brake
The older the driver, the higher the ratio
that could not recognize the warning.
Age
20 to 39
Elderly drivers have high motivation to drive.
40 to 59
60 to 69
0%
20%
40%
60%
80%
100%
Can be used anytime, anywhere
Drivers over age 60 recognized the
warning but did not brake.
Trains and buses cannot be used
1
Trains and buses limit time
Trains and buses take too much time
Fig. 1 Reaction of drivers to active safety system
(Experimental study using Toyota Driving Simulator)
Love to drive for fun
0%
10%
20%
30%
40%
50%
Fig. 2 Vehicle necessity
Integrated Safety
8/30
60%
70%
80%
90%
ADAS concept with driving intelligence
Shared control between an expert driver model and actual driver
Driver can override the system if necessary
•
Driver-vehicle cooperative system >= Experienced driver
Assistance system
•
Computer
Driver-in-the-loop ADAS concept (not fully automated)
Human driver
•
Integrated Safety
9/30
Haptic-based control is
important.
Safe driving concept: Obtaining anticipatory driving information
- Learn from the knowledge and experience of experienced drivers
Accident!
10 sec
1 sec
5 sec
(1) Determine how to drive
(2) Predict hidden risks
100 msec
10 msec
(3) Emergency avoidance
Collision avoidance
PCS
Damage mitigation PCS
Knowledge/experience-based information (mechanical learning) /
external information (maps, V2X)
Awareness of in-vehicle sensors
Higher sensor performance
Innovative sensing/recognition/judgment algorithms
Speed
Anticipatory
driving information
Prediction
Vehicle path
Forward distance
m
300
Integrated Safety
Intersection/
crossing
60
Stopped vehicle
Gradient
Tail of congestion
Road profile Traffic restriction Road friction
Merge
information
200
100
10/30
km/h
20
100
Control scheme of experienced driver model
Vehicle motion
Driving environment
Vehicle motion
Driver maneuver
Human driver
Digital map
Path/speed
modification
Objects
Location
Steering command
Accelerator command
Brake command
Actuator control unit
Radar
Shared control
Desired trajectory
+
Trajectory planner
Course tracking
Desired speed
+
Distance control
Safe speed control
Camera
Acc.
Brake
(1) Normal driving model
Risk potential prediction
Artificial intelligence
Human factors
Steer
Obstacle
Env. model
Acc. OFF
Knowledge-based DB
Brake ON
(2) Risk prediction model
(3) Emergency crash avoidance model
Vehicle speed
Vehicle position and heading
Integrated Safety
11/30
Driving intelligence: (1) Normal driving control
(1) Normal driver model for longitudinal and lateral direction, i.e. ,
car following and lane keeping, integrating the risk potential field
Road Boundary
Experimental Vehicle
Vehicle Trajectory
Predicted Vehicle Position
Target Path
Sensing the position of the driver’s vehicle using the curb
measured by laser radar, combined with GPS/digital map
Driver’s
vehicle
自車
Repulsive斥力ポテンシャル
force potential
Speed control using curvature and lane tracking
Proceeding
先行車 vehicle
k
Car following control using repulsive and attractive forces in the
risk potential field
Integrated Safety
12/30
Typical hazardous situation from near-miss database
Near-miss incident data collected in the real world can be used for the assessment
and advanced development of autonomous driving intelligence systems.
Integrated Safety
13/30
(2) Risk prediction model
Risk prediction model: Defensive driving, object motion prediction, hazard anticipation
Near-miss data
High risk
Low risk
Contour of risk potential generated from surroundings and on-road objects
Driving intelligence must predict the possibility of the appearance of pedestrians from
behind an object and determine the optimum safe speed.
Integrated Safety
14/30
Demonstration of driving with risk prediction
(1) Normal driving control of longitudinal and lateral dynamics,
i.e., car following and path tracking
(2) Risk potential based control including hazard anticipation to minimize collision risk
(3) Emergency avoidance control by braking and steering
Anticipatory driving intelligence is demonstrated using a Toyota Prius as shown below.
No pedestrian appears
(2) Risk potential based control
Integrated Safety
Pedestrian appears from
behind the parked car
(2)+(3) Emergency avoidance
15/30
Effect of defensive driving by driving intelligence
Longitudinal distance to Pedestrian [m]
• Collision avoidance performance can be effectively enhanced by
introducing risk prediction into the driver model (Fig. 2).
• The driving intelligence model can express expert driver behavior (Fig. 1).
Fig. 1 Comparison with data of expert drivers
20
(amax=2.5 m/s2)
(amax=7
● with risk prediction
▲ without risk prediction
m/s2)
15
10
Collision
unavoidable
5
0
0
10
20
30
40
Velocity [km/h]
50
Collision
avoidable
Fig. 2 Collision avoidance performance
Synthesis of defensive driving model by learning expert driver data for
enhancing driving safety.
Integrated Safety
16/30
(3) Emergency avoidance control: New Pre-Collision Safety (PCS) system
- Steering control added to conventional emergency brake.
- System controls the steering if a collision cannot be avoided by braking.
Pedestrian
Safe space
Avoidance by automatic
braking and steering
Integrated Safety
17/30
Conceptual diagram of ADAS with driving intelligence
Integrated Safety
18/30
Smaller 3D Lidar
CMOS detector
chip
SPAD Lidar*
Polygon
mirror
LASER
beam
Size: 1/6
LASER
emitter
Output by dist. image
Smaller/highly reliable/optimized processing
* SPAD: Single Photon Avalanche Diode, Lidar: LIght Detection And Ranging
Integrated Safety
19/30
Two topics for tomorrow
1. ADAS with Driving Intelligence for a Safer and
More Secure Traffic Society for Elderly Drivers
(S-Innovation Project in Japan)
2. Traffic Congestion and Highway Driving
Intelligence
Integrated Safety
20/30
Situation of highways in Japan
Congestion
Others: 2%
Accidents
*2008
* 2007
Other: 7%
Construction: 1%
Vs. person: 1%
Single vehicle: 14%
Accident:
29%
Sagtunnel:
53%
Other: 4%
Toll gate: 2%
Junction: 9%
Collision: 9%
Integrated Safety
Stopped car
in lane:
35%
Stopped car
at toll gate: 7%
Rear-end collisions: 67%
Traffic overconcentration: 68%
Targets
Moving car:
23%
Reduction of
- Traffic congestion
- Traffic accidents
- CO2
Global & local driving intelligence
 V2X, ACC/cooperative ACC
 Smart highway traffic system
21/30
V2V: Cooperative Adaptive Cruise Control (C-ACC )
 With millimeter-wave radar, vehicle to vehicle communication capabilities (760 MHz) have been
added to ACC to maintain an appropriate distance between leading and following vehicles.
 C-ACC provides a greater margin of safety for the driver, helps to prevent waves of deceleration,
and contributes to reduced highway traffic congestion.
V2V
Preview
control
Relative velocity & distance
control of ACC
Radar
Driving
force
Fig. 1 Control overview
Decelerating (90km/h  30 km/h, 0.2 G)
(1)
(2)
(3)
(4)
100
90
先頭車
トヨタ
スバル
マツダ
(1)
(2)
(3)
(4)
70
60
80
70
60
50
50
40
40
30
30
20
20
10
10
0
60
Amplifying
65
70
75
time[s]
先頭車
トヨタ
スバル
マツダ
90
車速[km/h]
80
車速[km/h]
Vehicle speed (km/h)
100
80
Human driver
85
0
60
90
Damping
65
75
time[s]
Time (sec )
ACC
Fig. 2 String stability performance of C-ACC (field operational test )
Integrated Safety
70
22/30
C-ACC
80
85
90
Traffic congestion reduction effect by simulation
Simulation to evaluate the quantitative relationship between the amplification rate and the penetration
rate. (e.g. amplification rate of 0.98 is needed to halve the congestion rate at a penetration rate of 20%)
Congestion
reduction rate
0.02 G
deceleration
Congestion
40 km/h
Amplification rate
Os2
=
of headway
Os1
Time headway to ensure a
basic traffic flow rate of more
than 2,000 veh/h depending on
the ACC penetration rate.
Penetration rate of ACC
Simulation result
Large effect on reducing congestion
0.9
0.7
0.5
Amplification rate of headway (ACC performance)
Congestion before widespread adoption of ACC - Congestion after widespread adoption of ACC
Traffic congestion
=
reduction rate
Congestion before widespread adoption of ACC
Integrated Safety
23/30
Automated highway driving
Automated Highway Driving Assist (AHDA)
Integrated Safety
24/30
Automated driving technology: Gate-to-gate
Integrated Safety
25/30
Summary: Cyber-physical system for intelligent driving systems
Big data
• Global driving intelligence
Near-miss data analysis/accident analysis
Enhanced map for dynamic usage
Traffic congestion prediction
etc.
• ICT: Info. & communication technology
V2X
Travel conditions database
Vehicle
Fｅｅｄｂａｃｋ
Human factor database
• Local driving intelligence
ADAS with autonomous driving intelligence
Accident
avoidance
Optimum
Fun to drive
eco-driving
Integrated Safety
26/30
Expectations for non-volatile memory
Advanced driver
assistance
Integrated vehicle
control
Calibration
Power train
Driving intelligence
Performance
Non-volatile
memory
 New NV-memory for next-gen systems
Integrated Safety
Vehicle
Computer
• Retention period: over 20 years
• Sleep & wake up power reduction
Automated
driving
- Localized map model
- Sensor fusion model
- Motion planning
- Driver model etc.
 Flash-memory for current systems
• Computing power (inst-fetch/data-ope)
Memory size
increasing
Machine learning
Reliability
• Writing times: over 105 times
20XX year
Connected
driving
Flash
Memory
Summit
2015
27/30
• Huge data storage for machine learning
- Localized map model, sensor fusion
model, motion planning, drive model, etc.
• Random access speed
• High reliability
Rewarded with a smile
by exceeding your expectations
Thank you for your kind
attention!
Integrated Safety
28/30