1. Proactive Traffic Merging Strategies for Sensor-Enabled Cars
Ziyuan Wang, Lars Kulik and Kotagiri Ramamohanarao
Department of Computer Science and Software Engineering
The University of Melbourne, Australia
VANET 2007, September, 2007

2. Outline Introduction Problem Statement Progress So Far
Future Directions

3. Traffic Congestion Some facts on traffic congestion
Total amount of delay: 3.7 billion hours in 2003
Wasted fuel: 2.3 billion gallons lost
Congestion cost: $63 billion
Source: Texas Transportation Institute, 2005 Urban Mobility Report.

4. Major Causes of Congestion
Bottlenecks:
Intersections of on-ramps and main roads
Blockage due to obstacles
“slinky type” effect
Source: Federal Highway Administration. Traffic Congestion and Reliability: Linking Solutions to Problems - Executive Summary.

5. Emergence of VANETs Sensor-Enabled Cars
Spatial information
Dedicated Short-Range Communications (DSRC)
Vehicle-to-Vehicle (V2V)
Vehicle-to-Roadside (V2R)
Vehicular Ad hoc Networks (VANETs)
Safety: less accidents
Efficiency: higher road utility
Position
Speed
Acceleration
Deceleration

6. Problem Statement Goal How Applications Optimize traffic throughput
Proactive traffic merging algorithms
Technology available: sensor-enabled cars + VANETs
Applications
Intersections at the ramp and the main road of highways (Highway merge assistant)
Lane changing when there are obstacles on the way

7. Existing Approaches Traffic signal timing Ramp metering
Fixed
Traffic-responsive
Ramp metering
Real-time information
Automation
Fully: Platoon (tightly grouped cars)
Partial: Adaptive Cruise Control (ACC)
Limitations
Adaptive
Flexible
Robust
Traffic conditions are highly variable and unpredictable

8. Contributions
Proposed proactive traffic merging algorithms that aim to use the current road facilities efficiently
Designed a controlled simulation environment intended to test various traffic merging strategies
Investigated what criteria are significant to evaluate the performance of traffic merging algorithms

9. Proactive Merging AlgorithmB X Y
Regular
Proactive
Highway bottleneck
Regular strategy
Local decision
Distance-based
Velocity-based
Regular means from normal traffic laws. There are different various version of laws, whatever law u follow, give ways to one stream, and always making decision at the merging point.
The figure shows a scenario where ramp cars merge onto a main road. In both cases we assume the same initial configuration (middle figure). In the upper figure we assume a priority-based merging algorithm where a car does not adapt its speed before it arrives at the merging section. This requires car x to slow down considerably in order to merge onto the main road. In the lower figure, we assume that car x adapts its speed before its arrival at the merging section and can merge immediately when it arrives at this section. This leads to a smaller impact on both traffic streams as the merging car has the same speed as the cars on the main road when it merges.

10. Outline of Our Algorithms
Comparisons of the proactive merging algorithms
Strategy
Information
Right of Way
Assumption
Distance-based
Position
The car that is closest to the merging point
Velocity does not vary much
Velocity-based
Velocity
The car that arrives to the merging point first
Acceleration does not vary much

11. Outline of Our Algorithms
Sliding decision point
Adjust speed appropriately
Input
{c, d, e}
{x, y}
Merging strategy
Output {c, d, x, e, y} {c, x, d, y, e} {x, c, d, y, e}
Regular
Distance
Velocity

12. Evaluation Metrics Delay Throughput Flow
The time to fill up the main road with a certain number of cars from the ramp
Throughput
The number of cars that complete merging over a period of time
Flow
The product of density and velocity

13. Simulation Intelligent Driver Model (IDM) Microscopic traffic model
Parameter
Value
Maximum velocity
Safe time headway
Maximum acceleration
Maximum deceleration
100 km/h
1.5 s
1m/s^2
3m/s^2
Intelligent Driver Model (IDM)
Microscopic traffic model
Safety distance
Exit ramp

14. Experiments and Results
Experiment settings
Light Medium Heavy
Unit
Main road
Ramp
cars/km ∞

15. Experiments and Results

16. Summary Traffic merging strategies benefit from sensor-enabled cars
Proactive merging algorithm outperforms regular strategy in terms of throughput and delay
Achieved at the cost of slightly lower velocity
Findings so far
We must manage complex tradeoffs among factors,
such as velocity, throughput, and latency.
The traffic flow increases at the beginning as the ramp cars merge, and decreases as the average velocity decreases. In order to maintain a relatively high traffic flow as well as a large number of cars get into the loop, we need to find a balance between the velocity and the number of cars merge into the loop.
Now I’m talking the immediate future work focuses on robustness.

17. Robustness of Algorithms
Human factors
Imperfect information
Sensor accuracy
Unreliable communication medium
Studies* show only 50-60% of cars in range will receive a car’s broadcast
Penetration rates
Initially, only a small number of sensor-enabled cars
Despite the human factors we talked before, some other factors of importance are involved in this issue.
Considering those factors will also enable a more realistic model.
* Source: J. Yin, T. EIBatt, and S. Habermas, Performance evaluation of safety applications over DSRC vehicular ad hoc networks, VANET 2004

18. Higher Degree of Realism
Obstacles
Blocking
Traffic patterns
Different distributions
Multiple lanes
Lane-changing
Heterogeneity
Different types of vehicles
We will address some other factors
Why Poisson

19. Thank you!
Questions,
Suggestions,
&
Comments