This action might not be possible to undo. Are you sure you want to continue?

# An Intelligent Coordinated Traﬃc Signal Control Based on EVALPSN

**Kazumi Nakamatsu1 , Jair M. Abe2 , and Seiki Akama3
**

University of Hyogo. Himeji, Hyogo Japan nakamatu@shse.u-hyogo.ac.jp 2 Paulista University. Sao Paulo, SP-Brazil jairabe@uol.com.br 3 Kawasaki. Kanagawa, Japan sub-akama@jcom.home.ne.jp

1

Abstract. We introduce a coordinated traﬃc signal control method based on EVALPSN. The simulation results of the EVALPSN coordinated traﬃc signal control method are shown with comparing to conventional ones. Keywords: EVALPSN(Extended Vector Annotated Logic Program with Strong Negation), traﬃc signal control, paraconsistent annotated logic program, defeasible deontic reasoning.

1

Introduction

We have proposed a paraconsistent annotated logic program called EVALPSN (Extended Vector Annotated Logic Program) that can deal with defeasible deontic reasoning and paraconsistency [1]. Some applications of EVALPSN to various kinds of control and safety veriﬁcation, robot action control, discrete event control, and safety veriﬁcation for railway interlocking and air traﬃc control have been introduced [3,4]. Traﬃc jam caused by inappropriate traﬃc signal control is a serious environmental issue that we have to overcome. Conventionally, traﬃc signal control is divided into three kinds, single, coordinated and broad area ones according to the number of traﬃc signals to be controlled. In single traﬃc signal control each traﬃc light is controlled independently. On the other hand, in coordinated one a series of traﬃc lights on the same route are controlled. Recently an optimizing method for traﬃc signal control by Genetic Algorithm has been proposed and applied practically. However, the method takes a long time to compute optimal solutions and is not so appropriate for real-time control. We have already proposed an intelligent real-time single traﬃc signal control system based on EVALPSN and shown that it could reduce the traﬃc density 10% to 15% by simulation [2]. In this paper, we extend the idea of EVALPSN single traﬃc signal

B. Apolloni et al. (Eds.): KES 2007/ WIRN 2007, Part II, LNAI 4693, pp. 869–876, 2007. c Springer-Verlag Berlin Heidelberg 2007

4 to detect traﬃc density or ﬂowing traﬃc. → all red → green → yellow → all red → T1. Based on this control policy the traﬃc signal is controlled in real-time by defeasible deontic reasoning in EVALPSN. 2. On the other hand.2. Then.3. 2 Traﬃc Signal Control in EVALPSN First of all we suppose a basic traﬃc signal control policy that traﬃc lights treating a larger amount of traﬃc should have the priority to be green until reducing the traﬃc density. J.1. Akama Control Direction ← S3 SC2 S2 T4 T2 T1 T3 S4 S1 SC1 Fig. Abe. In single traﬃc signal control.1 as the object of the single traﬃc signal control.2. and they have the following signal chart.1 Single Traﬃc Signal Control We take a series of three typical intersections in Japan as described in Fig. there is a conﬂict between the permission and forbiddance. Nakamatsu. We assume that: the middle intersection has four traﬃc lights T1.3.2 → red → red → all red → T3. The ﬁrst demand can be regarded as permission for turning the light from red to green and the second one can be regarded as forbiddance from turning the traﬃc light from green to yellow or red.870 K. you must have a demand for turning the traﬃc light from red to green in your mind. yellow and all red terms are ﬁxed . yellow and red.4 . if you are driving through the intersection with green you must have a demand for keeping the traﬃc light green. green.4 → green → yellow → all red → red → red the intersection has four sensors S1. 1.1. Then. each one has three kinds of displays. Intersections in Japan control to coordinated one and propose an EVALPSN based coordinated traﬃc signal control system. and S. The basic idea of EVALPSN traﬃc signal control is to resolve such conﬂicts by EVALPSN defeasible deontic reasoning. we focus on only the middle intersection among the three ones in Fig. Suppose that you are waiting for the traﬃc light turning from red to green at an intersection. which are described as white boxes in Fig.M. We will formalize the traﬃc signal control in defeasible deontic formulas ﬁrstly and translate them into EVALPSN later.

if one of the sensors S3. t). the permission for turning the traﬃc lights T3. Then. Basically. 1.q are called the prerequisites and consequents of the rules A → p and B ⇒ q.2 (t) ∧ T1. (1) rg where. the forbiddance from turning the traﬃc lights T3.2 (t) denotes that one of the sensors S1. we introduce defeasible reasoning brieﬂy. rg S1. rg S3. t) ⇒ ¬T3. For example. the minimum and maximum lengths of green term are given in advance. (2) .4 are red and green. In the EVALPSN traﬃc signal control.4 are red and green at the time t. thus.4 from green to yellow should be derived.2 (r. two kinds of rules. t) ∧ T3.4 from green to yellow should be derived. t) ∧ T3. If one of the sensors S1. and the conﬂict between the permission and forbiddance should be resolved by EVALPSN defeasible deontic reasoning. T1. Defeasible Deontic Rules for Single Traﬃc Signal Control. t) denote that the traﬃc lights is T1.4 (y. then the rule R2 defeats the rule R1 and only the consequent ¬p of the defeasible rule R2 is derived.4 (g. where A. t). if the sensor S1 detects ﬂowing traﬃc. the green term of the traﬃc light T1 should be extended within the range between the minimum and maximum terms. For example.4 detects the ﬂowing traﬃc over the criterion at the time t. The details are found in [5]. we will formalize the traﬃc signal control in defeasible deontic formulas and translate them into EVALPSN. t) ⇒ ¬ ¬T3. that is to say. 2. S1.4 (g.2 and T3. Firstly. and both a and b hold. the ﬂowing traﬃc or traﬃc density of each road connected to the intersection are regarded as the forbiddance or permission for turning the object traﬃc light from green to yellow. On the other hand. Defeasible reasoning is carried out based on the superiority relations between defeasible rules. respectively.2 (r. As the ﬁrst step. respectively. green term must be controlled between the minimum and maximum terms.B and p. Strict rules are used for deriving facts and superior to defeasible rules. Now we will start formalizing the single traﬃc signal control for the middle intersection in defeasible deontic formulas.An Intelligent Coordinated Traﬃc Signal Control Based on EVALPSN 871 and only green term is variably controlled according to the traﬃc detected by the sensors. and the symbol a modal operator to denote obligation. if there is the superiority relation R1 < R2. This condition is represented by the defeasible rule. Then.4 (y. the following conditions should be considered as strict or defeasible rules. a strict rule A → p and a defeasible rule B ⇒ q are used. respectively.2 and T3.4 (t) ∧ T1. the symbols ¬ ¬ and ¬ denote permission and forbiddance.2 detects the traﬃc density over the criterion at the time t.2 (r. Suppose that there are conﬂicting defeasible rules: R1: a ⇒ p and R2: b ⇒ ¬p. We suppose that the traﬃc lights T1. respectively. we have the defeasible rule conﬂicting to the defeasible rule (1). t) and T3.2 detects traﬃc density more than the criterion at the time t.4 (g.

4 should be guaranteed its minimum length. (4) There are conﬂicting defeasible rules (1) and (2) whose consequents are permission and forbiddance.2 is 2 hours but that of T3. t) → T2 (r. if the forbiddance is defeasibly derived at the time t.2.3. Minimum Green Term Rule. it is forbidden from the traﬃc light turning from green to yellow. Let M INi (g. t) → T1 (r. the defeasible rule (2) is considered to be superior to the defeasible rule (1). (5) (6) Moreover. if it is obligatory that the traﬃc light T1 is red at the time t. 3. the traﬃc light T1 must be red at the same time.4 (g. t) → T3. These conditions can be represented by the following strict rules : T3.3.4 (y.4 (g. This rule can be represented by the strict rules. On the other hand.4 (g. (3) Maximum Green Term Rule.4 (g. T3.4 also should have its maximum term. As the defeasible reasoning result.872 K. Such synchronization is represented by the following strict rules. If we take only the defeasible rules (1) and (2) into account. and if it is obligatory that the traﬃc light T3 is green at the time t. t) → ¬T3.. M IN3. Then we have a rule : if the green time is longer than its maximum term at the time t. which are represented by the following deﬁnite rules. 2. J. Each green term of the traﬃc lights T1. t) → ¬ ¬T3. the traﬃc light has to be turned from green to yellow at the next time t + 1. the signal has to keep the current state at the next time t + 1. we will translate the strict and defeasible rules into EVALPSN. t) ∧ T3. In order to avoid such an extreme control. t) ∧ ¬T3. t) and T3 (g. t) (7) EVALPSN for Single Traﬃc Signal Control.4 from green to yellow is prior to the permission for it. we need some more strict rules for synchronizing all the traﬃc lights. 4}) denote that the green time of the traﬃc light Ti has not passed its minimum term at the time t. t + 1). t). t + 1). (i ∈ {1.4 (y. Then we have a rule : if the green time is shorter than its minimum term at the time t. t) ∧ ¬ ¬T3. 4}) denotes that the green time of the traﬃc light Ti has already passed its maximum term at the time t. that is to say. Akama We also have to deﬁne the superiority relation (<) between the defeasible rules (1) and (2).4 (y.4 (y. t) ∧ T3. As we assume that maintaining the current signal state is superior to turning it to the next signal state.4 (g. Nakamatsu.M.4 (g. we have to control green term in an appropriate range. The strict and defeasible rules can be . For example. Now. We need more traﬃc signal control rules. if the permission is derived at the time t. T1 (r. Abe. t). t). 3. Let M AXi (g. the traﬃc light T2 also must be red at the same time. we might have an extreme situation such that the green term of T1.4 (g. M AX3.4 (y. This rule can be represented by the strict rules.4 is 1 minute. (i ∈ {1. t). it is deﬁnitely permitted for the traﬃc light turning from green to yellow. and S. Each green term of the traﬃc lights T1. 2.2. the forbiddance from turning the traﬃc lights T3. t) → T3.

1). γ] → T3.4 (y. β].2 are red and the traﬃc lights T3.4 are green. β].4 are forbidden for turning to yellow.4 (g. the minimum green term has already passed. Case-1. α] ∧ ∼ M AX3.1 as an example. t + 1) : [(2. t) : [(2. 0).4 (y. 2. and the rest of the sensors do not detect then.4 (g. (12) T3. (9) The EVALPSN clause (8) represents that: if the traﬃc sensors S1 or S2 have detected the traﬃc density over the criterion.2 (r. 0).4 are green. t + 1) : [(2. The defeasible rules (1) and (2) are translated into the EVALPSN clauses. α] → T3. β] is derived.4 (y. α] → T3.4 (g.4 (g. t) : [(0. t) : [(2. 1). If one of the sensors S1. α] ∧ T3. 0). and the maximum green term has not passed yet under the same condition that the traﬃc lights T3. 0). α] ∧ T3. t) : [(0.4 (g. furthermore. (10) M AX3. α] ∧ T3. 0). 1). 0). Suppose that the traﬃc lights T1. 1). Similarly. α] ∧ T3. γ]. t) : [(2.4 (g. γ] is derived.4 (y. the EVALPSN clause (12) is also ﬁred and the obligation T3. t) : [(2. t + 1) : [(2. α]∧ ∼ S rg 4 (t) : [(2.4 (y. 0). then the traﬃc lights T3.4 (t) : [(2. β] → T3. t) : [(2.4 (y. On the other hand.2 (t) : [(2. 0). furthermore. and the traﬃc sensors neither S3 nor S4 has detected the ﬂowing traﬃc over the criteria at the time t under the condition that the traﬃc lights T3.4 (y.4 (g. 0). 2).4 (g.4 are permitted for turning to yellow. γ]. t) : [(0.2 (r. 0). 0). α] ∧ ∼ M IN3. 0). β].4 (y. 0). the EVALPSN clause (8) is ﬁred and the permission T3. are also translated into the EVALP clauses.4 (g.4 (y. it has already passed the minimum green term then. t) : [(2. t) : [(0. If both the sensors S1 and S3 detect the traﬃc density and ﬂowing traﬃc over the criteria at the time t. the EVALPSN clause (13) is also ﬁred and the obligation T3. t) : [(2. the EVALPSN clause (9) represents that: if the traﬃc sensors S3 or S4 have detected the ﬂowing traﬃc over the criterion.2 Coordinated Traﬃc Signal Control We show the basic idea of the EVALPSN coordinated traﬃc signal control with taking the same intersections in Fig. furthermore. (11) T3. α] ∧ T1.4 (g. α]∧ ∼ S rg 3 (t) : [(2. β]. 0). t) : [(0. the strict rules (3). S rg 3. 0). α] → T3.4 are green. t) : [(0. t) : [(2. α] ∧ T3. 2). then the traﬃc lights T3. (4). respectively.4 (y. t) : [(2. 1). 0). 1). Case-2. t) : [(2. (5) and (6). t) : [(2.2 detects the traﬃc density over the criterion at the time t and the rest of the sensors do not react then. 0). Here we focus on only the traﬃc from the right to the left and its converse direction as the object of the . 0). t + 1) : [(2.An Intelligent Coordinated Traﬃc Signal Control Based on EVALPSN 873 translated by formalizing their semantics in EVALPSN. β] is also derived. α] → T3. S rg 1. (13) (8) Example 1. M IN3. the EVALPSN clause (9) is ﬁred and the forbiddance T3. 0). t) : [(0. 0). t) : [(0. β] is also derived.4 (g.4 (g. α] ∧ T1. t) : [(2. α] ∧ T3.

4 (y. Nakamatsu. that is to say. 0). t) : [(0. not only each term of green. 0). α] ∧ ∼ S rg 4 (t) : [(2. α] ∧ ∼ M IN3. Scrg (t) ∧ T1.2 (r.(2) < (14). α] ∧ T3. we have the defeasible rule. (14) where Scrg (t) denotes that one of the sensors Sc1.2 and T3. α]∧ ∼ Scrg 1 (t) : [(2. γ].4 (t) : [(2.2 is processed. 1). If one of the sensors Sc1 or Sc2 detects ﬂowing traﬃc over the criteria at the time t..2 . We assume the same condition as the single traﬃc signal control. β]. On the other hand.4 (g. Abe. Then. those defeasible rules are retranslated into the EVALPSN clauses.4 from green to yellow should be derived. t) ⇒ ¬ 1. t) : [(2. only the green term of each coordinated traﬃc light should be controlled by EVALPSN defeasible deontic reasoning based on traﬃc amount information. the permission for turning the traﬃc lights T3. t) : [(2. Therefore.e. α]∧ ∼ Scrg 2 (t) : [(2. J. Akama EVALPSN coordinated traﬃc signal control. we have the following conditions in defeasible deontic formulas. t) : [(0. 0). Then.2 detects the ﬂowing traﬃc 1.2 described by black boxes in Fig. In conventional coodinated traﬃc signal control. (16) (15) . As the coordinately controlled route traﬃc from the right to the left or its converse are regarded to be superior to other traﬃc. that is to say the traﬃc lights T1. 0).874 K. and S. 3. 0).2 (r. the permission for turning the traﬃc lights T3. 0). 0). respectively. 1). Defeasible Deontic Rules for Coordinated Traﬃc Signal Control.4 (y. α]∧ ∼ S rg 3 (t) : [(2. (1) < (2) < (14) into account.. t). we note that when the traﬃc information detected by the remote sensors Sc1. yellow and red but also a timelag called an oﬀ set between two neighbor traﬃc lights are controlled. the extra traﬃc sensors Sc1. 0). t) : [(2.e. t) : [(2.2 (t) : [(2. t) ∧ T3. which is detected by traﬃc sensors attached to not only the object intersection but also its both neighbor intersections.2 over the criterion.2 (r. α] ∧ ∼ M AX3. (2) and (14). t) : [(2. Moreover.4 (g. α] ∧ T3.4 (y. i. S rg 1. and they should be preinstalled. α] ∧ T1. 0).2 ¬T3. α] → T3. the defeasible rule (14) is considered to be superior to the defeasible rule (2). i. EVALPSN for Coordinated Traﬃc Signal Control. in EVALPSN coordinated traﬃc signal control. 0). 0). S rg 3. when we consider the EVALPSN coordinated traﬃc signal control in terms of the middle intersection. Taking the superiority relations among the defeasible rules (1).M. We also need to consider the superiority relation between the conﬂicting defeasible rules (2) and (14). appropriate timelags between the detection and processing of the traﬃc information should be taken into account according to the distances between the object intersection and the sensors Sc1. 0). α] → T3.4 are red and green.4 (g.4 (g.4 from green to yellow has a prior to the forbiddance from it.1 attached to the right and left intersections should be taken into account to detect the traﬃc ﬂowing into the middle intersection. α] ∧ T1. t) : [(2.4 (g.

If one of the sensors Sc1. γ] is derived. (11). therefore. the EVALPSN clause (12) is also ﬁred and the obligation T3. the green term is 60 steps. where 15% probability means 15 cars/100 steps appear at the road. t) : [(0. the EVALPSN clause (17) is ﬁred and the permission T3. 0).4 (g.2 and Sc1. t + 1) : [(2. respectively. the EVALPSN clause (17) is ﬁred and the permission T3. t) : [(2. (17) As well as the EVALPSN single traﬃc signal control the coordinated one also has Minimum and Maximum Green Term Rules. and EVALPSN coordinated ones for the intersections in Fig. Example 2.An Intelligent Coordinated Traﬃc Signal Control Based on EVALPSN 875 Scrg 1. Case-5. the green term is between 14(min) and 60(max) steps. cars are ﬂowing into all the intersections from each road in 10% probability except for the traﬃc from the right road.4 (y. t + 1) : [(2. α] ∧ T3. (12). Case-3. α] ∧ T1. We have assumed the following conditions for simulation: the unit time called “step” is deﬁned in the simulation system. EVALPSN single. γ] is derived. 0). If all the sensors S1. conventional (ﬁxed time) single. both the EVALPSN clauses (15) and (17) are ﬁred and the same permission T3. 0). “travel step” denotes the total number of steps . α] ∧ ∼ M IN3. 0). β] is also derived.4 (y.2 detect the traﬃc density and ﬂowing trafﬁc over the criteria at the time t. t) : [(0.4 (y.2 detect the ﬂowing traﬃc or trafﬁc density over the criteria at the time t. (13) as the translation of those rules. for EVALPSN traﬃc signal control. γ] is derived. 0). and there are 5000 steps/hour in fact.“stop step”denotes the total number of steps that the 50 cars stopped.4 (g.2 (t) : [(2.4 and Sc1. furthermore. which is 15%. respectively. 1).4 (y. t) : [(0. furthermore. furthermore. t) : [(0. t + 1) : [(2. All the four kinds of simulation have been carried out during 5000 steps each and repeated ten times. respectively.2. 0).2 (r. 0). each distance between the three intersections is the same 20 cells. we also have the EVALPSN clauses (10). the EVALPSN clause (12) is also ﬁred and the obligation T3.4 (y. t) : [(2.1 by the cellular automaton simulation method. 1). and the oﬀ-set in the coordinated traﬃc signal control is 24 steps. t) : [(2. for ﬁxed-time traﬃc signal control.4 (y. If the sensors S1. Case-4.4 (y. We took the simulation data of 50 cars that are chosen at random among all the cars ﬂowing into the intersections from the right and traveling to the left.3. γ]. β] is also derived. the EVALPSN clause (12) is also ﬁred and the obligation T3. the yellow term is 3 steps. which is the time that a car travels one cell to the next cell. and the yellow term is ﬁxed 5 steps. β] is also derived. 3 Simulation We present the simulation results of the four kinds of traﬃc signal control. Suppose the same conditions as Example 1. coordinated. α] → T3. where “total car” denotes the total number of cars that ﬂowed into all the three intersections from any road.1 show the average values of the ten times simulation results. 1).2 and the sensor S4 detect the ﬂowing traﬃc over the criteria at the time t. 1). The simulation data in Tab.

: Annotated Semantics for Defeasible Deontic Reasoning. Oxford Science Publications. Nute. Nakamatsu. Szczuka. M. References 1. J. W.. Heidelberg (2003) 3. Mita. Abe.. 719–723.: A Discrete Event Control ´ Based on EVALPSN Stable Model Computation.M. 4 Conclusion We have introduced a coordinated traﬃc signal control method based on an annotated logic program EVALPSN and its computer simulation results by the cellular automaton method. J. vol. vol. K. A.. S-L. and S. lots of sensor installation are required and too much cost. Q. 2005.F. Simulation Results Fixed-time Single Fixed-time Coordeinated EVALPSN Single EVALPSN Coordinated total car 3804 3823 4149 4169 stop step 66 65 43 34 travel step 177 173 152 122 travel car 495 500 577 607 for the 50 cars having traveled from the right to the left. 671– u 681. Nakamatsu.M. pp. In: Ziarko. J. LNCS (LNAI). In: Slezak. pp. LNCS (LNAI)... Intelligent Automation and Soft Computing 13. T. H. Y. In: del Cerro. Nakamatsu.) Intensional Logics for Programming. LNCS (LNAI). Springer. Abe. “travel car” denotes the total number of cars that traveled from the right up to the left. 3641.. Springer. Heidelberg (2005) 4.. (eds. since each traﬃc light is controlled independently as one autonomous agent in the EVALPSN coordinated traﬃc signal control. Skowron..M. D¨ ntsch. and especially EVALPSN coordinated control is most eﬃcient among them. pp. Shibata. Chung.. Yao. Oxford (1992) . K. Abe. Y. D. Komaba. K. EVALPSN control can be implemented in both existent software and hardware such as PLC(Programmable Logic Controller). it is more fault tolerant than conventional coordinated one. vol. A. Y. (eds. Wang. 2639.. This simulation results clearly say that: EVALPSN traﬃc signal control shows more eﬃcient results than conventional ﬁxed time one in both single and coordinated traﬃc signal control methods. 126–154. Heidelberg (2001) 2. M.: An Intelligent Action Control System Based on Extended Vector Annotated Logic Program and its Hardware Implementation. Yao. D. Yao. (eds. (eds.) RSCTC 2000. Suzuki. 470–478. Seno. K. Suzuki..: Intelligent Real-time Traﬃc Signal Control Based on a Paraconsistent Logic Program EVALPSN. G. Nakamatsu.: Basic Defeasible Logic. Liu.. L. Nakamatsu...) RSFDGrC 2003..) RSFDGrC 2005.876 K. G. The EVALPSN traﬃc control is a sensor based realtime one and more eﬃcient than conventional ones from the simulation results. 289–304 (2007) 5. pp. In: Wang. although it has not been addressed in this paper. I.. Penttonen. A. if it is implemented practically.. Suzuki.. Springer. however. Akama Table 1. T. A. Y..