Paper Review:Automated On-Ramp Merging for
Congested Traffic Situations
Emmanuel Sean Peters
Objectives & Results
Develop an automated merging system that:
I. Permits merging traffic to fluidly enter the major road to avoid congestion on the minor road
II. Modifies the speed of vehicles on main road to minimize the effect on the already congested main road
A fuzzy controller and decision algorithm that uses Vehicle-to-Infrastructure communication is designed and tested using three production vehicles
Objectives & Results
Solved one of the major causes of congestion in urban environments: merging from minor to major roads.
• validated using simulations and real experiments involving mass produced vehicles
• system successfully & safely merged a car from ramp onto the main road at low speeds
Outline
I. Introduction
II. Control Architecture
III. Description of Vehicles & LCS
IV. Automated Ramp Entrance System
V. Experiments
VI.Conclusion
Outline: Introduction
• Increase in number of drivers and vehicles and cars on road over past few decades
• Urban environments are most congested• Advanced Driver-Assisted Systems (ADAS)– Ultimate decision is the driver’s; driver may be
incorrect
• Simulations are encouraging but…– Gasoline-propelled vehicle dynamics at very low
speeds are highly non-linear and difficult to model
Outline: Architecture
AUTOPIA Program:
The development of automatic cars using mass produced vehicles and tests on real roads
Outline: Vehicles & LCS
Automated Vehicles- 2 Citroёn C3s (Gasoline-Propelled)- 1 Berlingo Citroёn (Electric)
Local Control Station (LCS) - Detecting risky traffic situations & advising the vehicles
involved
Outline: Automated Ramp Entrance System (I)
Design of system divided into 3 phases
I. Detection
II. Optimal Merging Algorithm
III. Intelligent Controller Design – uses reference data from optimal merging algorithm
Outline: Automated Ramp Entrance System (II)
Decision System
- Ensures sufficient headway is achieved by the time the merging point is reached
Control System
- Fuzzy logic
- Relies on SpeedError & Distance Error values
Outline: Experiments
Details: Decision System (I)
Details: Decision System (II)
Details: Control SystemFuzzy Logic- Solutions based on vague information- Mamdani Inference: max-min method- Membership functions – maps input a value between 0 and 1
Inputs: - SpeedError- DistanceError
Output:- Pedal [-1,1]
Weights:- Throttle – 40%- Break – 10%
Details: Control System
Details: Control System
DistanceError
Difference between leading
& trailing vehicles’ speeds
SpeedError
Difference between leading
& trailing vehicles’ speeds -
[-3,3]kmh
Three membership functions & three linguistic labels - Positive linguistic used to accelerate/brake for SpeedError/DistanceError - Negative linguistic used to brake/accelerate for SpeedError/DistanceError - Center linguistic used to indicate that trailing car maintaining target speed or distance
Details: Control System
Output variable, Pedal [-1, 1], as a function of fuzzy input variables SpeedError and Distance Error; determines which actuator is pressed
.
Details: Control System
Output variable as Sugeno singletons
Details: Simulation Results
L=10m, initial speed=3m/s
Scenario 1: x20=-18 and x30=-8, Scenario 2: x20=-26 and x30=-16
Details: Simulation Results
Scenario 1 Scenario 2
Details: Simulation Results
Scenario 1 Scenario 2
Details: Experiments
http://www.iai.csic.es/autopia/Videos/Merging.wmv
Questions?