Problem Description

Objectives

1. Minimize delays, stops, trip times, queue size
2. Maximize 'bandwidths'

Planning Measures

1. Optimize intersection timing
   1. Adding more sensors
   2. Single intersection/multiple (costly/noisy communication model?)
2. Public transit
   1. Fares
   2. Routes and schedules
3. Road tolls
   1. Parking restictions
4. Route and departure guidance

Commuter Decisions

1. Trip/Route planning
2. Mode selection

Software:

GIS Sources

• Stats-Can Maps (both GML and SHP formats)

Simulators

Commercial

• VISSIM / VISUM (http://www.ptv-ag.com/cgi-bin/index.pl)
• TRANSIMS (http://transims.tsasa.lanl.gov/), this is a traffic model developed by physicists at the Los Alamos, and sold to IBM
• Paramics (http://www.paramics-online.com/projects.php)
• CORSIM (http://mctrans.ce.ufl.edu/featured/tsis/version5/corsim.htm)
• LISA+ (http://www.schlothauer.de/en/lisa.html)
• Simtraffic (http://www.trafficware.com/)
• Aimsun (http://www.aimsun.com/site/)
• MECCA-UTS

Traffic Adaptive Control Systems

• ALLONS-D - online and distributed, arbitrary-cycle, tree-search optimization, local/regional. UMich.
• SPOT/UTOPIA - online and distributed, arbitrary-cycle, local/regional. Designed by the FIAT Research Centre
• OPAC - online and distributed, arbitrary-cycle, local/regional
• SPPORT - online and distributed, arbitrary-cycle, local
• PRODYN - online and distributed, arbitrary-cycle, local
• SCOOT - online and centralized, fixed-cycle, regional
• TRANSYT - offline and centralized, fixed-cycle, regional
• SCATS - offline and centralized, fixed-cycle, regional
• Webster's Equisaturation - offline and centralized, fixed-cycle, regional

People

• Ying Zhang PARC, Traffic Sensor Networks
• Peter Stone CPSC @ UTexas at Austin
• Kurt Dresner CPSC @ UTexas at Austin
• David Pritchard CIVIL MSc at UToronto, extensive bibliography
• Jacqueline Jenkins CIVIL @ UBC
• Nicolas Saunier CIVIL Post-Doc @ UBC
• Lawrence Frank SCARP @ UBC
• Andrew Nash Independent Consultant
• Nicholas V. Findler EECE @ Arizona State University

Papers

Traffic Light Optimization

Title	Author	Scope	Problem	Solution	Comments
Adaptive Look-Ahead Optimization of Traffic Signals	Porche & Lafortune	Single intersection, extension to multiple	Minimize total delay	Build an MDP, tree-search (Branch and Boundy)	Good overview and taxonomy of different adaptive controllers. Presents ALLONS-D.
Optimizing Networks of Traffic Signals in Real Time - The SCOOT Method	D Robertson and R Bretherton	multiple intersections, human-controlled vehicles	To minimize queues and vehicles stops at intersection	SCOOT improved on TRANSYT (signal coordination based on fixed time plans) (1) measure CFP in real time (2) update online model of queues continuously (3) incremental optimization of signal settings	Definitely a useful real scenario on signal coordination technique
A Distributed Approach for Coordination of Traffic Signal Agents	Ana L. C. Bazzan	Multiple intersections	Maximize throughput and traffic saftey; minimize travel times and environmental costs	Pseudo-agent multiagent learning approach	Some good cites. The multiagent learning seems a little hacky
Game Theory: Potential Applications in Transportation Planning	Karim A. S. Ismail	Multiple intersections	Finding out applications for game theory in traffic	Driver timing, Road Pricing, Traffic Assignment, Public Transport, Signal Timing	Looks like a good start
Traffic adaptive control of a single intersection: A taxonomy of approaches	R.T. van Katwijk, B. De Schutter, and J. Hellendoorn	Single Intersection	Minimize the delay experienced by vehicles through manipulation of the traffic signal timings.	Various, mostly MDP-type frameworks. Dynamic programming and tree-search used to optimize	Good overview of adaptive controllers, but Porche & Lafortune (1997) is a better written overview
Multi-Agent Reinforcement Learning for Traffic Control	Marco Wiering	2x3 intersection	Optimize Traffic lights	Simlulate with discrete simulation; use RL
Effects of Co-Evolution in a Complex Traffic Network	Bazzan et al.	6x6 grid	Optimize lighting control	Look at Greedy or Adaptive drivers, Greedy, Adaptive, or Q-Learning lights.	Small slice of agent, relatively few iterations allowed + no burn-in
Reinforcement Learning-based Control of Traffic Lights in Non-stationary Environments	Oliveira et al	9 intersection grid	Optimize light policy	Model with cellular automata, use RL learning

Other

Title	Author	Scope	Problem	Solution	Comments
Traffic Intersections of the Future	K Dresner and P Stone	Single intersection, automated drivers	Increased efficiency and throughput	Tile-based Reservation	Good references
Using Intelligent Agents for Urban Traffic Control Systems	D Roozemond	Multi-Intersection, human-controlled vehicles	General optimizing framework	Rule-based pseudo-agent approach	Terrible paper, no real content
10 Years with LHOVRA - What are the experiences	P Engstrom	multiple intersections	Reduce stops, delays, and accidents in rural & urban areas	L=Truck Priority, H=Major road priority, O=Incident reduction, V=Variable green/yellow, R=Reduction of red light infringement	Seems to be poorly translated, I didn't get much out of this paper.
Different Policy Objectives of the Road Pricing Problem – a Game Theory Approach	D Joksimovic et al	one origin-destination (OD) pair, 2 possible routes, 2 travellers, 1 person as road authority	(road pricing) model as 3 types of games: Cournot, Stackelberg and social planner game	For different objectives, multiple optimal solutions exist. the objective functions may have a non-continuous shape	toy problem but interesting, definitely could be extended
Fictitious play for finding system optimal routings in dynamic traffi�c networks	A Garcia et al	some simulated traffic network with travellers	(dynamic traffic assignment) minimize average trip time experienced in the network	use repeated play of fictitious games that eventually weakly converge to a local system optimal routing (Alliance software)	"assuming minimizing average travel time is common interests of all travellers" is a little too simplistic.
Optimal Dimensions of Bus Service Zones	S K Chang and P M Schonfeld	urban area divided into bus service zones	(public transport system) optimize design of urban bus service zones by jointly optimizing decision variables (service headway, route spacing, route length, demand density)	analytic optimization model	optimization based on static data - i guess they can't change bus routes all the time, lots of references about optimizing design of transit systems
Passenger Arrival Rates at Public Transport Stations	Marco Luethi, Ulrich Weidmann, Andrew Nash	Zürich	How does one model passenger arrival statistically, given bus frequency	Uniform distribution with shifted Johnson distribution	Summarizes some previous approaches to modeling arrival rates. Worried about overfitting, and how general this result is (data collected in Zürich)
Traffic Calming in Three European Cities: Recent Experience	Andrew Nash	Three European Cities	n/a	n/a	A laundry list of how Zurich, Vienna, and Munich have dealt with traffic calming
Learning Cooperative Lane Selection Strategies for Highways	D. Moriarty and P Langley	Single highway	How should 'smart cars' switch lanes in a highway?	Evolutionary algorithms	Strange problem setting (a device that tells you what lane you should be in, once you give it a speed preference?), strange solution
Implementation of the OPAC adaptive control strategy in a traffic signal network	Nathan H. Gartner et al.	Multiple intersection	Minimize delays and stops	Rolling horizon tree-search approach, which synchronization layer	Seems like a good approach, similar to ALLONS-D in goals and motivations. Gartner has other publications, and seems to cite few other people than himself... draw your own conclusions :)
An Adaptive Interactive Agent for Route Advice	S Rogers, C Fiechter, P Langley	route advice system	How to build an adaptive interactive agent to generate route advice based on user preferences	Agent generates route choices and update user model by observing feedback from user, assign costs to attributes of roads (travel time, length, road type)	pretty simple approach, using some kind of learning/optimization, possible future work include better learning algorithm + taking into account of dynamic attributes (current road conditions)
A Collaborative Driving System Based on Multiagent Modelling and Simulations	S Halle, B Chaib-draa	some automated cars on a straight, one way, two lanes, highway segment	which are good coordination models for collaborative driving system (platoons of collaborating vehicles)	compared to centralized model, the multiagent teamwork model is more safe and flexible, but requires more messages to be communicated.	part of the Automobile of the 21st Century (Auto21) supported by Government of Canada, interesting idea, try to tackle traffic from driving perspective instead of from redesigning traffic facilities
The Network Effects of Alternative Road Pricing Systems	A D May et al	network analysis done for city of Cambridge and York	which 4 charging systems perform better in terms of its positive impact on traffic distribution	network analysis produced different results than conceptual analysis, possible future work include incorporating dynamic route guidance so that drivers are aware of the possible charges in advance	good facts about realistic charging systems for road use - make me wonder how highway 407 in Toronto does the charging.
Title	Author	Scope	Problem	Solution	Comments
Title	Author	Scope	Problem	Solution	Comments