A METHODOLOGY FOR TRAFFIC SIGNAL
CONTROL BASED ON LOGIC PROGRAMMING
Giovanni Felici
Istituto di Analisi dei Sistemi ed Informatica (IASI-CNR),
Consiglio Nazionale delle Ricerche
Giovanni Rinaldi
Istituto di Analisi dei Sistemi ed Informatica (IASI-CNR),
Consiglio Nazionale delle Ricerche
Antonio Sforza
Dipartimento di Informatica e Sistemistica,
Università degli studi di Napoli Federico II
Klaus Truemper
Department of Computer Science
University of Texas at Dallas
1
Outline of presentation
• Logic programming for traffic control
• The application
• Performance Evaluation
• Detectors Data
• Floating Probe Car
2
FACTS about Traffic Control:
• Small adjustments of the length of the phases ( 5 to10 secs) can
produce consistent savings
• Signal synchronization can be driven by traffic in a decentralized
fashion
• The control system must be able to adapt to irregular intersections
• The control system must learn as it works
• Traffic detection is crucial. There is a trade-off between quantity and
quality of the information, and it is important to find the right balance
for each intersection.
3
Research Project initially funded by
Progetto Finalizzato Trasporti 2 - CNR:
•
•
•
•
•
Istituto di Analisi dei Sistemi ed Informatica (IASI - CNR)
University of Texas at Dallas, Computer Science Program
Centro Studi sui Sistemi di Trasporto (CSST Roma)
project started in 1993
use of state of the art tools for Logic Programming and Logic
Optimization (the Leibniz System)
• use of a visual traffic microsimulator to implement and test different
control strategies
control strategies developed by this tool have proved to generate
consistent savings when compared with traditional traffic control
systems
4
Main features of Control System
Adaptive
Decentralized
Based on
a Logic
Model
The control decisions
depend on the state of the
current traffic.
Traffic detection and
decision making are
performed in real time
Better use of the available resources
Reactions to fluctuations in traffic
flow
Each signal is controlled by
an independent control unit.
No supervision is needed.
Neighboring units exchange
a
limited
amount
of
information
Low cost hardware
No fixed-charge installation
Modularity
Reliability
The state of the traffic, the
decision variables, and the
control
strategies
are
expressed in first order
logic
Easy to understand
Can reproduce human expertise
Extremely flexible
Readily modeled by traffic engineer
5
Traffic Variables
Decision Variables
Control Strategy
The state of traffic at the proximity of the intersection is detected by a
set of traffic detectors and is translated into True/False values of
logic predicates
The decisions are represented by logic variables associated with
transitions between the phases
The control strategy is represented by a set of logic statements
that connect traffic and decision variables using the Leibniz Syntax
6
Visual Microsimulation
 Micro Traffic Simulator for urban networks:
 Each car is simulated independently with car-following principles
 Each signal is simulated
 Several traffic generation patterns
 Traffic behaviour and effectiveness of logic strategies can be visually
evaluated
 Statistics on performance indicators and traffic patterns can be collected
7
A Simulated Session
8
Network design
Control
strategy design
Visual test
Logic algorithm
compilation
Performance
analysis
 Network of Workstation Unix
 C standard language with X11 graphic libraries
 Distributed computation over more workstations for real time simulation
 Built-in Leibniz interface
9
THE APPLICATION: Afragola
Partners:
• IASI-CNR (Istituto di Analisi dei Sistemi ed Informatica)
• TechNapoli consortium
• Dipartimento di Informatica e Sistemistica, Università degli studi di
Napoli Federico II
• ELASIS, Sistema Ricerca FIAT nel mezzogiorno
• CSST Napoli (Centro Studi sui Sistemi diTrasporto, FIAT)
•
•
University of Texas at Dallas, Department of Computer Science
Tecnosistem
•
SelfSime (Signal Control Hardware)
10
Main characteristics of the installation:
•
Autoscope Camera detection system :
• 5 presence counters and 3 queue counters for each approach (4)
• 2 cycles, one with 2 and one with 4 phases
• traffic detected is often noisy or not precise due to the position of the
cameras; also the topology of the intersection makes virtual loops fail
at times
•
The control system receives data from the detectors and produces the
control decision (switch to next phase or stay in current phase) every 3
seconds
•
The Logic Strategy:
• 104 logic variables
• 185 logic statements
• max solution time below 0.05 second
11
Performances Evaluation
•
3 different control methods were tested on the same intersection:
• fixed time where fixed cycle was obtained with TRANSYT
• dynamic adaptive system built-in in Selfsime signal hardware
• logic control
•
Performances compared by:
• data from detectors
• floating probe car
12
Evaluation: Data from Detectors
•
Indicator: sum of occupancy figures of all queue counters
•
comparisons are made for similar traffic conditions
•
we consider comparisons of two methods only if experiments were
run on the same day, same hour, and same incoming traffic
(tolerance of approx. 5%)
•
very good behaviour of logic control just by observation
•
logic control is consistently better than fixed time and dynamic control
13
30,00
25,00
20,00
15,00
10,00
5,00
0,00
10,30-16
11-12
12,01-13
13,01-14
14,01-15
15,.01-16
Experiments
LOGIC CONTROL VS. DYNAMIC: PERCENTUAL SAVINGS
15,00
10,00
5,00
15
,.0
116
14
,0
115
13
,0
114
12
,0
113
11
-1
2
10
,3
016
19
,0
120
18
,0
119
17
,0
118
16
,0
117
16
,0
120
16
,3
017
,0
3
15
,1
016
14
,1
014
,4
5
-5,00
12
,1
513
0,00
10
-1
1
percentage
percentage
LOGIC CONTROL VS. FIXED TIME: PERCENTUAL SAVINGS
Experiments
14
Floating Probe Car
15
Floating Probe Car
•
14 paths around the intersection
•
round trip time
•
average speed
•
fuel consumption
•
emission of HC and CO
16
AFRAGOLA
17
PATHS 1, 2, 4, 14
18
PATHS 6, 7, 9, 10
19
PATHS 3, 5, 8, 11
20
PATHS 12, 13
21
POINTS MAPPED ON THE GIS – GPS ERROS
22
POINTS MAPPED ON THE GIS –ERRORS CORRECTION
23
POINTS MAPPED ON THE GIS – CORRECTED PATHS
24
Floating Probe Car
SEGMENT
1
2
3
4
average
TIME ON SEGMENTS
DYNAMIC
LOGIC
53,33
64,38
97,56
43,17
50,58
48,50
74,00
82,00
68,87
59,51
SAVINGS
20,7%
-55,8%
-4,1%
10,8%
-13,6%
25
Floating Probe Car
STOPS ON SEGMENTS (< 10kmh)
SEGMENT
DYNAMIC
LOGIC
SAVINGS
1
9,69
12,26
26,5%
2
16,78
5,67
-66,2%
3
6,45
4,00
-38,0%
4
11,58
13,69
18,2%
average
1
2
3
4
DYN
12,51
17,42
12,97
14,67
FUEL
LOG
14,51
11,99
12,21
15,71
SAV
16%
-31%
-6%
7%
average
14,39
13,60
-5%
SEGMENT
11,13
DYN
2,59
3,73
2,65
3,07
HC
LOG
3,04
2,44
2,49
3,25
SAV
17%
-34%
-6%
6%
3,01
2,81
-7%
8,90
-20,0%
DYN
43,22
68,54
44,76
54,07
CO
LOG
53,23
40,05
41,09
57,94
SAV
23%
-42%
-8%
7%
52,65
48,07
-9%
26
Scarica

Floating Probe Car