Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 8, No. 1, February 2018, pp. 246 253 ISSN: 2088-8708 246       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     T raffic Light Signal P arameters Optimization Using Modification of Multielement Genetic Algorithm I Gede P asek Suta W ijaya 1 , K eeichi Uchimura 2 , and Gou K outaki 3 1 Informatics Engineering Dept., Engineering F aculty , Mataram Uni v ersity , Indonesia 2,3 Electrical Engineering and Computer Science Dept., K umamoto Uni v ersity , Japan Article Inf o Article history: Recei v ed: Jun 3, 2017 Re vised: No v 29, 2017 Accepted: Dec 14, 2017 K eyw ord: Artificial intelligence GA optimization signal parameters transportation system ABSTRA CT A strate gy to optimize traf fic light signal parameters is presented for solving traf fic con- gestion problem using modification of the Multielement Genetic Algorithm (MEGA). The aim of this method is to impro v e the lack of v ehicle throughput ( F F ) of the w orks called as traf fic light signal parameters optimization using the MEGA and P ar ticle Sw arm Opti- mization (PSO). In this case, the modification of MEGA is done by adding Hash-T able for sa ving some best populations for accelerating the recombi nation process of MEGA which is shortly called as H-MEGA. The e xperimental results sho w that the H-MEGA based opti- mization pro vides better performance than MEGA and PSO based methods (impro ving the F F of both MEGA and PSO based optimization methods by about 10.01% (from 82,63% to 92.64%) and 6.88% (from 85.76% to 92.64%), respecti v ely). In addition, the H-MEGA impro v e significantly the real F F of Ooe T oroku road netw ork of K umamoto City , Japan about 21.62%. Copyright c 2018 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Name I Gede P asek Suta W ijaya Af filiation Informatics Engineering Dept., Engineering F aculty , Mataram Uni v ersity Address Jl. Majapahit 62 Mataram, Lombok, INDONESIA Phone +62-37-636126 Email gpsuta wijaya@unram.ac.id 1. INTR ODUCTION The traf fic congestion is big problems which causes man y ne g ati v e ef fects not only to road users ph ysiological b ut also t o economic and en vironmental [ 1]. Ph ysiologically , the traf fic congestion mak es the pedestrians and dri v ers ha v e to pay a lot of attentions during on the roads. Economically , the traf fic jam increases the fuel consumption, which implies to transportation cost. En v i ronmentally , the traf fic jam increases the pollution of v ehicle disposa l g as such as C O 2 raising the greenhouse ef fect on the en vironment. There are three cate gories of strate gy to opt imize traf fic signals which are w ork ed based on the le v el of v ehicle in v olv ement [2]. The first cate gory utilizes le g ac y de vices with no v ehicular in v olv ement, which can be to redefine the signal timing of the junction using certain technique. The second cate gory utilizes v ehicles on the road to wirelessly transmit data about themselv es (e.g. location, v elocity). It means the signal timing is optimized by considering the reports of v ehicles on the roads. The last cate gory seems costly because it requires sophisticat ed de vices and softw are to performing automatically the optimization on-board. In this research, the first cate gory of traf fic light signal parameters optimization is proposed by modifying the Multielement Genetic Algorithm using Hash-T able which is shortly called as H-MEGA. The H-MEGA is an impro v e- ment of pre vious w orks called as traf fic light signal parameters optimization using the Multielement Genetic Algorithm (MEGA) and P article Sw arm Optimization (PSO) [1, 3]. 2. RELA TED W ORKS Some w orks for traf fic light signal parameters optimizations ha v e been proposed which can be classified to three approaches: firstly , using artificial intelligence (GA, Fuzzy , Neural Netw orks) and their v ariations; secondly , using statistical such as stochastic[4]; and finally using v ehicle in v olv ement [2]. Among them, the approaches using J ournal Homepage: http://iaescor e .com/journals/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v8i1.pp246-253 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 247 artificial intel ligence play important roles for traf fic light signal parameters optimizations such as approaches based on PSO [1], GA [5, 6, 7, 8, 9], fuzzy logic which determine the best signal parameters using fuzzy rule [10, 11]. Ho we v er , some of them were not implemented on real road netw ork and had lack of performances. T raf fic light optimization also can be performed by considering v ehicular in v olv ement via communication de vices. The Ref. [12] also de v eloped a signal control algorithm that allo ws for v ehicle paths and si gnal control to be jointly optimized based on adv anced communication technology between approaching v ehicles and signal controller . Ho we v er , the algori thm assumed that v ehicle trajectories could be fully optimized and it w as de v eloped assuming a simple intersection with tw o single-lane. The Ref. [13] proposed signal setting optimization on urban road transport netw orks which w ork ed based on tr a v el demand to congested road transport netw ork. In this case, tw o interacting pro- cedures are de v eloped to solv e the system of models: (i) an optimization procedure to obtain an optimal configuration of signal setting parameters and (ii) an assignment procedure, incorporating a path choice model with e xplicit path enumeration and a flo w propag ation model, to capture the ef fects of signal setting configuration on user path choice be- ha vior . The Ref. [14] presented traf fic bot tleneck identification and optimization. T w o main f actor traf fic bottlenecks are signal timing at intersections together with static properties of left-turn and straight-through lanes of roads[14]. The ant colon y algorithm w as proposed to find out optimal coordinated signal timing for a re gional netw ork. The Ref. [15] had proposed an optimization of pedest rian phase patterns and signal timings for isolated intersection which establishes quantitati v e criteria for selecting pedestrian phase pat terns between the e xclusi v e pedestrian phase (EPP) and the normal tw o-w ay crossing (TWC) with both safety and ef ficienc y f actors traded-of f in an economic e v aluation frame w ork. The proposed method is able to assist transportation professionals in properly selecting pedestrian phase patterns at signalized intersections. The Ref. [9] also proposed intersection signal control multi-objecti v e optimization using GA, which w orks to obtain a signal control multi-object optimization method to reduce v ehicle emissions, fuel consumption and v ehicle delay simultaneously at an intersection. Moreo v er , the v ehicle anti-collision alert system in FPGA has been de v eloped to decrease the number of road accidents[16] which not only cause injurie s, deaths b ut also traf fic jam. It means the alert system is an de vice that can be used to drop-of f traf fic congestion. In addition, a v ariation of GA such as optimization using MEGA has been proposed for finding traf fic light signal parameters[3, 7, 17]. That method has been pro v ed to solv e traf fic congestion in real Ooe T oroku road netw ork, K umamoto Shi, Japan. Ho we v er , it is lack of netw ork throughput (percentage of v ehicle flo w) and time consuming on obtaining the optimal traf fic light signal parameters on the Aimsun 6.1 for simple road netw ork (see Fig. 3(a)). T o impro v e MEGA s performance, particle sw arm optimization (PSO) w as emplo yed instead of MEGA[1]. Ho we v er , it just impro v ed 3.13% of MEGA s achie v ement. In addition, it also needed almost the same computational time. 3. THE OPTIMIZA TION ALGORITHM The optimization algorithm is based on H-MEGA that is emplo yed to search the optimum of fset, c ycles, splits time of four nodes/junctions of Ooe T oroku road netw ork. The Ooe T oroku road netw ork (Fig. 3(b)) is located in K umamoto City Japan, at latitude and longitude point 32.81 and 130.72 or in url: https://www.google.com/ maps/@32.8054628, 130.7218806, 17z . It is one of road netw ork ha ving most traf fic congestion in K umamoto city . The properties of Ooe T oroku road netw ork including the node/j unction, signal model, and signal timing has been clearly presented by Ref. [1]. 3.1. T raffic Light Signal P arameters Each node/junction has traf fic light equipped with signal parameters: of fset, c ycle, Y ello w , all Red, and split [3, 7]. The of fset parameter is the time coordination between traf fic light (node) representing the starting of green signal timing. F or instance, the Node 1 and 2 ha ving 0 and 3 seconds of fset parameters means that the Node 2 starts the c ycle signal timing at 3 seconds after started the c ycle of Node 1. The c ycle parameter represents the total time of traf fic light starting from Green and returning to Green. The Y ello w and all Red are usually defined constantly representing the duration of yello w and all red signal of the traf fic light. The split that consists of main and sub split means the Green time percentage of main road and sub road, respecti v ely . In this paper , the optimization algorithm searches the optimum of fset, c ycles, and split to get maximum v ehicle gone out, minimum v ehicle in and w ait out, less v ehicle stop, and short delay time of considered real road netw ork. 3.2. Modification of MEGA There are some v ariations of genetic algorithm (GA) which were de v eloped to solv e specific problems[3, 18, 19]. F or instance, the multielement GA (MEGA[3]) w as de v eloped to optimize traf fic light signal parameters, the parallel GA[18] w as de v eloped for solving the uni v ersity scheduling problem, and the augmented GA[19] w as formed to utilize feature reduction on data mining. The algorithm of MEGA for finding the best traf fic light signal parameters T r af fic Light Signal P ar ameter s Optimization Using Modification ... (I Gede P asek Suta W ijaya) Evaluation Warning : The document was created with Spire.PDF for Python.
248 ISSN: 2088-8708 is gi v en in Fig. 1(a). In MEGA, the populations consist of man y chromosomes e xtracted from the road netw ork traf fic lights. The MEGA, which is also included by elitism, has been pro v ed that it could be used to find good traf fic light signal parameters as presented in Refs. [7, 8, 17]. In this paper , the MEGA and PSO based opt imization are impro v ed by modifying the MEGA using Hash- T able (H-MEGA). This idea comes from the PSO algorithm which w as inspired by social beha vior of bird flocking or fish schooling[20]. It means the solution is searched in around current optimum solution. Therefore, a Hash-T able ha ving k e y for inde xing the n -best populations is added to MEGA. Lik e PSO, the best solution of MEGA is just searched in around the Hash-T able by performing the recombination such as selection, crosso v er and mutation. The dif ferent between MEGA and H-MEGA is presented in Fig. 1. There are some addition processes to impro v e the MEGA (Fig. 1(a)) which are sho wn by light green block (see Fig. 1(b)) as follo ws: 1. H-MEGA initialization which gi v es initial v alue Hash-T able size, number of populations and chromosomes. 2. Putting first n -best fitness to Hash-T able means first n -best populations corresponding to first n best fitness are appended to Hash-T able for ne xt recombination process. The recombination process in v olving selection, crosso v er , and mutation are performed based on the populations e xisted in Hash-T able. 3. Deleting the same populations: populations result of recombination that are same as e xisted populations in Hash- T able are deleted for decreasing the computation time because the y pre viously ha v e been e v aluated. ( a )   M E G A ( b )   H - M E G A S t a r t I n i t i a l i z e   ( M E - G A ) F o r G e n = 0 t o N - 1 d o C a l c u l a t e   F i t n e ss  ( P o p   [ i ] ) S o r t i n g   F i t n e ss Cro s s o v e r Se l e c t i n g Pa re n t   Po p u l a t i o n D e f i n i n g c u t - p o i n t D o i n g t h e   c ro s s o v e r R o u l e t t e   W h e e l F o r   se l e c t i o n M u t a t i o n D e f i n i n g t h e   g e n   f o m u t a t i o n D o i n g m u t a t i o n G e n = M a x ? G e n + + T h e   b e st P o p E n d D o i n g   S i mu l a t i o n   o n A P I   +   A I M S U N 6 . 1 N o Y e s G e n = M a x ? T h e   b e st P o p E n d N o Y e s D D e e l l e e t t e e Upda t e d P o p I n i t i a l i z e   H a sh   a n d   M E G A D o i n g   S i mu l a t i o n   o n A P I   + A I M S U N 6 . 1 F o r G e n = 0 t o N - 1 d o C a l c u l a t e   F i t n e ss  ( P o p   [ i ] ) S S o o r r t t i i n n g g F F i i t t n n e e s s s s P P u u t t t t i i n n g g f f i i r r s s t t - - n n b b e e s s t t F i t n e ss t t o o H H a a s s h h - - T T a a b b l l e e C C r r o o s s s s o o v v e e r r a a n n d d M M u u t t a a t t i i o o n n U U p p d d a a t t e e d d P P o o p p i i s s e e x x i i s s t t i i n n H H a a s s h h - - T T a a b b l l e e ? ? M M a a x x ? ? Y Y e e s s N N o o G G e e n n + + + + S S t t a a r r t t R R o o u u l l e e t t t t e e W W h h e e e e l l Figure 1. Flo w Chart of MEGA[3] and H-MEGA. In this case, the fitness formula for performing populations e v aluation is gi v en by F p = exp V w o C w o + exp V in C in + exp t 0 D C tD . Where the t 0 D is defined as t 0 D = t D tot D T r . The constant v alues ( C w o , C in , and C tD ) are gi v en as follo ws: C w o = 100, C in =500, and C tD =500. These constant v alues were chosen to minimize the ef fect of each v ariables to the fitness v alue. These v alues ha v e been utilized to e v aluate the PSO[1] and MEGA[8], and the y could obtain good solution. The parameters (v ehicle w ait out ( V w o ), v ehicle in ( V in ), tra v el distance ( tot D T r ), and time delay ( t D )) are tak en from simulation outputs. The V w o means the total v ehicles which are w aiting to enter into the road netw ork, V in means the total v ehicles that still e xist in the road netw ork, the tot D T r is total tra v el distance of the v ehicles in simulation, and the t D is defined as the delay time of v ehicles in simulation. 3.3. Optimization Pr ocess Optimization process in v olv es Aimsun 6.1 simulator , application interf ace (API) which is a DLL modul that is pro vided by Aimsun 6.1 written in C++, and H-MEGA modules. Aimsun 6.1 simulator is transport modeling soft- IJECE V ol. 8, No. 1, February 2018: 246 253 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 249 w are which is used to perform the traf fic simulati on of Ooe T oroku road netw ork. Aimsun 6.1 simulator is de v eloped and mark eted by TSS-T ransport Simulation Systems and is widely used by uni v ersities, consultants, and go v ernment agencies w orldwide for transportation planning, traf fic simulation, and emer genc y e v acuation studies[1]. It is em- plo yed to i mpro v e road infrastructure, reduce emissions, cut congestion and design urban en vironments for v ehicles and pedestrians. The coordination and communication of t hree modules of optimization process w orks based on Diagram block are gi v en in the Fig. 2. The optimization process can be described as follo ws: 1. The Aimsun 6.1 gi v es the API initial data for n populations, m -generations, yello w time, all red time, and the range v alue of of fset, c ycles, and split, 2. The API passes the initial data to H-MEGA and orders the H-MEGA performing initialization n -populations of traf fic light signal parameters. 3. Through the API, the n -populations of traf fic light signal parameters are sa v ed as output by H-MEGA . 4. The API orders the Aimsun 6.1 performing traf fic simulation on road netw ork for all n -populations of traf fic light signal parameters and sa v e the simulation results on the database. 5. After finishing traf fic simulation, the API passes results to H-MEGA for performing e v al uation and recombina- tion of all traf fic light signal parameters, and finally sa ving the results as ne w traf fic light signal parameters. 6. Repeat the point 3 to 5 until reaching m -generations.   AP I   H - MEGA   A I MS UN  6 . 1   S i m u l a t i o n   Out pu   V go V in V wo t D i s t &   t D e l ay   O u t p u t   H - M EG A :   S i g n a l   P a r a me t e r s   Figure 2. Diagram block of coordination and communication among Aimsun 6.1, API, and H-MEGA[1]. 4. EV ALU A TION AND DISCUSSION In order to kno w the performance of H-MEGA for obtaining the optimum traf fic light signal parameters, some e xperiments were carried out using tw o road netw orks: simple road netw ork (Fig. 3(a)) and real road netw ork (Fig. 3(b)). The e xperiments in the simple road netw ork w as to find out whether the H-MEGA can deli v er optimum traf fic light signal parameters. While the e xperiments in the real road netw ork w as to confirm that the H-MEGA could be used to find out the best traf fic light signal parameters. All e xperiments used 5 minutes w arming up and signal parameters constraints as follo ws: firstly , 0 O f f set 120 and O f f set = 1 ; secondly , 60 C y cl e 180 and C y cl e = 5 ; and thirdly , 10 S pl it 90 and S p l it = 5 .   Red  Nu m b e r :  Ro a d  I D   288   (a) Simple (b) Ooe T oroku Figure 3. T w o road netw orks for e xperiments[3, 7, 8, 17] T r af fic Light Signal P ar ameter s Optimization Using Modification ... (I Gede P asek Suta W ijaya) Evaluation Warning : The document was created with Spire.PDF for Python.
250 ISSN: 2088-8708 4.1. Experiment on Simple Road Netw ork Experiments on simple road netw ork were carried out using tw o netw ork states ha ving v ehicle flo w (VF) 4800 per hour which its distrib ution is sho wn in T able 1. The first netw ork state had straightw ay and turn left, while the second netw ork state had straightw ay and turn right signals[3]. The v ehicle turning percentage of each junction for simple netw ork is 50%. The first e xperimental results sho w that the proposed method can find the best traf fic T able 1. V ehicle flo w distrib ution in simple road netw ork model[3, 8]. Road  ID*   2 86   29 2   29 8     3 02     3 1   31 2     3 2   32 T otal  VF  8 00   40 0   40 0     8 00     8 0   40 0     4 0   80 48 0 * :   Ro ad   ID   o f   F i g .   3 ( a)  T able 2. Throughput of simple road netw ork model. No   P a t t er n   VF   M et ho d   V go   V in   V wo   =V F - V go   Dela y   T im e   F F ( %)   1   T h First  Netw o r k   S tate   4800   ME GA   4431   231   299   369   1 8 3 . 8 0   8 9 . 3 2   P SO   4308   242   427   492   1 9 9 . 1 5   8 6 . 5 6   H - M E G A   4449   266   256   351   1 9 9 . 9 4   8 9 . 5 0   2   T h Seco n d     Netw o r k   State   4800   ME GA   3269   347   1422   1531   4 7 0 . 9 0   6 4 . 8 9   P SO   3408   421   1183   1392   4 6 0 . 9 7   6 8 . 0 0   H - M E G A   3509   368   1 139   1291   4 7 4 . 4 5   6 9 . 9 6   T able 3. The ef fect of nElites to H-MEGA on the second netw ork states of simple road netw ork. P a t t er n   M et ho d   nE lite s   V go   V in   V wo   D   Dela y   T im e   F F ( %)   T h First  Netw o r k   State   H - ME GA   2   4449   266   256   351   1 9 9 . 9 4   8 9 . 5 0   4   4447   256   258   353   1 9 3 . 5 3   8 9 . 6 4   6   4467   251   238   333   1 9 4 . 0 4   9 0 . 1 3   8   4437   254   272   363   2 0 0 . 7 1   8 9 . 4 0   10   4411   290   258   389   1 9 7 . 0 7   8 8 . 9 5     light signal parameters for both teste d netw orks state. In addition, the proposed method pro vided almost similar performance as tw o most related methods (MEGA[3, 7, 8] and PSO[1], which is sho wn by almost similar throughput ( F F ) about 89.50% for the first netw ork state and 69.96% for the second netw ork state (See T able 2). It means that H-MEGA method is pro v ed that it can be emplo yed to search the best traf fic light signal parameters for solving the traf fic congestion on the simple road netw ork The second simulation w as carried out to kno w the ef fect of number of best populations on finding the best traf fic light signal parameters on simple road netw ork. In this simulation, the number of best populations (nElites) sa ving in Hash-T able w as v aried from 2 10 . The simulation results sho w that the best nElites for H-MEGA to search the best traf fic light signal parameters is six (6) which can pro vide the highest F F among the others, as sho wn in T able 3. In addition, this simulation result also sho ws that H-MEGA pro vides higher F F compared to that of MEGA and PSO of pre vious e xperiments (see T able 2). It confirms that the H-MEGA can be empl o yed to obtain the best traf fic light signal parameters of simple road netw ork. Re g arding to computational time, the H-MEGA needs much shorter computational time (41.73 minutes) than that of MEGA and PSO (62.54 and 50.85 minutes, respecti v ely). The computational time is defined as a total time that is required by Aimsun 6.1, API, and H-MEGA to accomplish the simulation with 40 populations and 50 generations. Mostly computational ti me in the simulation is influenced by Aimsun 6.1 which tak es the almost 0.671 s econds to simulate replication of road netw ork for 1 hour v ehicles mo v ement in the road netw ork. It means that the H-MEGA not only impro v e the traf fic light signal parameters b ut also the com pu t ational time of the e xisting methods. It can be achie v ed because the H-MEGA searches optimum t raf fic light signal parameters in the entire some best populations sa v ed in Hash-T able and the H-MEGA also does not performance the e v aluation on populations which are the same as those of in the Hash-T able. 4.2. Experiment on Real Ooe T or oku Road Netw ork Further e v aluation of H-MEGA w as carried out in rea l Ooe T oroku road netw ork (see Fig 3(b)). The Ooe T oroku road netw ork had 5510 v ehicles flo w (VF) per hour , which were distrib uted as presented in T able 4. It also had 3708 pedestrian flo w per hour that were distrib uted into four junctions/nodes: 636, 1386, 415, and 860 people for node 1, 2, 3 and 4, respecti v ely . The turning percentage of VF per hour of each road in Ooe T oroku road netw ork w as set by real data that were obtained from the Oee T oroku site which were manually counted at peaks sessions (8:00 AM to 9:00 AM)[1, 8]. In this sim u l ation, the H-MEGA w as compared to the related w orks: MEGA (base-line)[8] and PSO[1]. The fitness of population e v aluation during the simulation sho w that all methods tend to find best traf fic light signal parameters of real road netw ork, as presented in the Fig. 4. The H-MEGA tends to gi v e better performance in terms F F than MEGA and PSO, because the fitness of H-MEGA is sm aller than that of the others. F actually , by using the best traf fic light signal parame ters for Ooe T oroku netw ork obtained by H-MEGA (presented T able 5), the F F of H-MEGA (92.64%) is much higher than that of MEGA (82.63%) and PSO (85.76%) while the real F F is about 71.02% (see T able 6). From T able 6, the proposed method can impro v e significantly the traf fic congestion of real Ooe T oroku road netw ork by about 21.61% of the real F F . While the MEGA and PSO can impro v e by about 11.60% and 14.74% of IJECE V ol. 8, No. 1, February 2018: 246 253 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 251 the real F F . This simulation results are inline to simple road netw ork achie v ement. It reconfirms that the H-MEGA not only can obtain the best traf fic light signal parameters for solving the traf fic congestion b ut also can impro v e the performances of MEGA and PSO for real Ooe T oroku Road Netw ork. 1 100 10000 1 0 0 0 0 0 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 Fi t t ne s s   Ge ne ra ti ons   M E GA P S O H -ME GA Figure 4. Fitness of H-MEGA compared to e xisting methods. T able 4. The v ehicles flo w of Ooe T oroku road netw ork[1, 7, 17]. S h e e t 1 P a g e   1 V e h i c l e s F l o w / h ou r  of  R oad  ID * T o t al 297 298 304 307 310 314 316 319 C a r 486 586 1594 1122 318 164 432 456 5158 B us 8 18 24 30 0 0 2 12 94 T r uc k 24 28 64 64 12 6 30 30 258 *:   R o a d   I D   o f   Fi g .   3 ( b ) T able 5. The best traf fic light signal parameters of Ooe T oroku obtained by H-MEGA. Ju n c ti o n s   Of f se t   (s)   Cy c le   (s)   M a in   S p li (% )   S u b     S p li (% )   No d e   1   0   135   65   35   No d e   2   13   95   45   40   No d e   3   86   175   70   30   No d e   4   14   70   60   30     The traf fic congestion condition of before and after optimization were v erified by simulating the Ooe T oroku road netw ork using the original traf fic light signal parameters[1] and the best one (T able 5). The simulation results were compared in Fig. 5, which sho w that the real traf fic congestions are happen in the 8 roads singed by red roman number (see Fig. 5(a)).The hea vy traf fic congestions are happen in road section I, II, III, V , VII and VIII which is indicated by man y v ehicles queue symbolized by small blue rectangular . Ho we v er , when using the best traf fic light signal parameters, the traf fic congestions decrease significantly , as sho wn in Fig. 5(b). In det ail, hea vy traf fic congestions are just happen in road section I and VI. It is still happen because the v ehicles flo w from in the section I (Road ID 316 and 317 (Fig. 3(b)), is high enough 432 with road width just 3 meter which mean the road density is o v erflo w . From this v erification, the traf fic congestion can be solv ed by resetting the traf fic light signal parameters using appropriate ones which can be searched by artificial intelligence such GA, PSO, Neural Netw ork, etc. Ov er all, this v erification supports the pre vious conclusion that the H-MEGA is alternati v e solution for searching the optimum t raf fic light signal parameters and it also can impro v e the performances of MEGA[8] and PSO[1]. T able 6. Throughput of H-MEGA on Ooe T oroku road netw ork compared to mostly related w orks. N o   M e t ho ds   V e hi c l e   V F   V g o   V in   V w o   = V F- V g o   D e l a y   T i m e   D T r t o t ( k m F F ( % )   1   Re a l   Bu s   9 4   8 1   2 0   3 2   1 3   N A   N A   7 1 . 0 2   Ca 5 1 5 8   3 0 8 5   1 1 3 2   1 3 4 0   2 0 7 3   T ru c k   2 5 8   1 5 9   5 5   6 5   9 9   Pe d e s t r i a n   3 7 0 8   3 2 3 9   3 4   0   4 6 9   T o t a l   9 2 1 8   6 5 6 4   1 2 4 1   1 4 3 7   2 6 5 4   N A   2   Ba s e     L i n e [8 ]   Bu s   9 4   8 6   1 2   5   8   9 1 3 . 9 9   5 7 9 0 . 4 3   8 2 . 6 3   Ca 5 1 5 8   3 7 8 4   8 9 4   3 8 1   1 3 7 4   T ru c k   2 5 8   2 5 4   3 7   2 3   4   Pe d e s t r i a n   3 7 0 8   3 4 5 3   1 6 2   7 9   2 5 5   T o t a l   9 2 1 8   7 5 7 7   1 1 0 5   4 8 8   1 6 4 1   0 . 2 2 2 *   3   PSO [ 1 Bu s   9 4   8 2   7   5   1 2   1 6 8 2 . 4 2   6 0 8 9 . 8 3   8 5 . 7 6   Ca 5 1 5 8   4 0 5 2   7 4 2   3 2 9   1 1 0 6   T ru c k   2 5 8   2 7 3   5 0   1 7   - 1 5   Pe d e s t r i a n   3 7 0 8   3 5 0 2   1 5 0   1 3   2 0 6   T o t a l   9 2 1 8   7 9 0 9   9 4 9   3 6 4   1 3 0 9   0 . 2 1 2 *   4   H- M EGA   Bu s   9 4   9 7   6   0   - 3       7 9 5 . 2 0     6 9 6 1 . 8 1   9 2 . 6 4   Ca 5 1 5 8   4 5 4 3   5 2 8   4 5   6 1 5   T ru c k   2 5 8   2 9 4   2 9   3   - 3 6   Pe d e s t r i a n   3 7 0 8   3 6 2 3   6 9   0   8 5   T o t a l   9 2 1 8   8 5 5 7   6 3 2   4 8   6 6 1   0 . 0 9 3 *   Note:  *   The  delay  ti m div id e d   by   V go     5. CONCLUSION AND FUTURE W ORKS The proposed traf fic light signal parameters optimization using H-MEGA has been implemented successfully to find the best traf fic light signal parameters, which is sho wn by higher throughput of both simple and real road T r af fic Light Signal P ar ameter s Optimization Using Modification ... (I Gede P asek Suta W ijaya) Evaluation Warning : The document was created with Spire.PDF for Python.
252 ISSN: 2088-8708 (a) Real traf fic light signal (b) The best traf fic light signal of H-MEGA Figure 5. T raf fic congestion v erification of Ooe T oroku road netw ork us ing Aimsun 6.1 simulator using real and best traf fic light signal parameters. netw orks. In detail, the H-MEGA can increase significantly the throughput ( F F ) of real Ooe T oroku road netw ork by about 21.62% (from 71.02% to 92.64%). It means, the H-MEGA is successfully to search the best traf fics light signal parameters of considered junctions that af fects the decrease traf fic congestion on the Ooe T oroku road netw ork. In terms of computational time, the proposed method needs much shorter time for accomplishing the sim u l ation among mostly related methods (MEGA and PSO). In future, the Aimsun 6.1 simulator will be modeled by Neural Netw ork for decreasing the computat ional time of accomplishing the simulation. In addition, the proposed methods will be formulated for finding the best traf fic light signal parameters on comple x road netw ork . A CKNO WLEDGMENT W e w ould lik e to s end our great thank to Japan Students Services Or g anization (J ASSO) for funding of my research in GSST -K umamoto Uni v ersity . REFERENCES [1] I. G. P . S. W ijaya, K. Uchimura, and G. K outaki, “T raf fic light signal parameters optimization using particle sw arm optimization, in 2015 International Seminar on Intellig ent T ec hnolo gy and Its Applications (ISITIA) , May 2015, pp. 11–16. [2] R. Florin and S. Olariu, A surv e y of v ehicular communications for traf fic signal optimi zation, V ehicular Communications , v ol. 2, no. 2, pp. 70–79, 2015. [Online]. A v ail able: http://www .sciencedirect.com/science/ article/pii/S2214209615000121 [3] I. G. P . S. W ijaya, K. Uchimura, G. K outaki, T . Nishihara, M. S., S. Ishig aki, and H. Sugita n i , “The impro v ement me g a based traf fic signal control optimization using ne w fitness model, J ournal on Computing , v ol. 2, no. 2, pp. 64–69, 2012. [4] X. Y u and W . Reck ert, “Stochastic adapti v e control model for traf fic signal systems, T r ansportation Resear c h P art C , v ol. 14, pp. 263–282, 2006. [5] L. Singh, S. T ripathi, and H. Arora, “The optimization for traf fic s ignal control using genetic algorithm, Interna- tional J ournal of Recent T r ends in Engineering , v ol. 2, no. 2, pp. 4–6, 2009. [6] S. T akahashi, H. Kazama, T . Fujikura, and H. Nakamura, Adapti v e search of an optimal of fset for the fluctuation of traf fic flo w using genetic algorithm, IEEJ T r ans. on Industry Applications , v ol. 123, no. 3, pp. 204–210, 2003. [7] T . Ni shihara, N. Matsumura, K. Kanamaru, I. W ijaya, G. K outaki, K. Uchimura, H. Sugitani, and S. Ishig aki, “The v erification with real-w orld road netw ork on optimization of traf fic signal parameters using multi-element genetic algorithms, in Pr oceeding of 19th ITS W orld Congr ess , Austria, 2012. [8] I. G. P . S. W ijaya, K. Uchimura, G. K outaki, S . Ishig aki, and H. Sugitani, “Fitness e v aluation of multi element genetic algorithm for traf fic signal parameters optimization, in Pr oceeding of 3r d International Confer ence on Soft Computing , Intellig ent System and Information T ec hnolo gy , Bali Indonesia, 2012, pp. 58–64. [9] Z. Zhou and M. Cai, “Intersection signal control multi-objecti v e optimization based on genetic algorithm, J ournal of T r af fic and T r ansportation Engineering (English Edition) , v ol. 1, no. 2, pp. 153–158, 2014. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S2095756415301008 [10] C. Chou and J. T eng, A fuzzy logic controller for traf fic junction si gnals, Information Sciences , v ol. 143, pp. 73–97, 2002. IJECE V ol. 8, No. 1, February 2018: 246 253 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 253 [11] S. Rahman and N. Ratrout, “Re vie w of the fuzzy logic based approach in traf fic signal control prospects in saudi arabia, T r anspn Sys Eng and IT , v ol. 9, no. 5, pp. 58–70, 2009. [12] Z. Li, L. Elefteriadou, and S. Ranka, “Signal control opti mization for automated v ehicles at isolated signalized intersections, T r ansportation Resear c h P art C: Emer ging T ec hnolo gies , v ol. 49, pp. 1–18, 2014. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S0968090X14002939 [13] F . A. Marcian, G. Musolino, and A. V itetta, “Si gnal setting optimization on urban road transport netw orks: The case of emer genc y e v acuation, Safety Science , v ol. 72, pp. 209–220, 2015. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S0925753514001866 [14] S. Y uan, X. Zhao, and Y . An, “Identification and optimization of traf fic bottleneck with signal timing, J ournal of T r af fic and T r ansportation Engineering (English Edition) , v ol. 1, no. 5, pp. 353–361, 2014. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S2095756415302816 [15] W . Ma, D. Liao, Y . Liu, and H. K. Lo, “Optimization of pedestrian phase patterns and signal timings for isolated intersection, T r ansportation Resear c h P art C: Emer ging T ec hnolo gies , no. 0, pp. –, 2014. [Online]. A v ailable: http://www .sciencedirect.com/science/article/pii/S0968090X14002460 [16] A. Z. Jidin, L. S. Li, and A. F . Kadmin, “Implementation of algorithm for v ehicle anti-collision a lert system in fpg a, International J ournal of Electrical and Computer Engineering (IJECE) , v ol. 7, no. 2, pp. 775–783, 2017. [17] I. G. P . S. W ijaya, K. Uchimura, G. K outaki, and T . Nishihara, “T raf fic signal control optimization using genetic algorithm and signaling model modification, in Pr oceeding of 20th ITS W orld Congr ess , T ok yo, Japan, 2013. [18] A. Berisha, E. Bytyc ¸ i, and A. T ershnjaku, “P arallel genetic algorithms for uni v ersity scheduling problem, Inter - national J ournal of Electrical and Computer Engineering (IJECE) , v ol. 7, no. 2, pp. 1096–1102, 2017. [19] S. Si v asankar , S. Nair , and M. Judy , “Feature reduction in clinical data classification using augmented genetic algorithm, International J ournal of Electrical and Computer Engineering (IJECE) , v ol. 5, no. 6, pp. 1516–1524, 2015. [20] J. K ennedy and R. Eberhart, “P article sw arm optimization, in Pr oceedings of IEEE International Confer ence on Neur al Networks , Piscata w ay , NJ, 1995, p. 19421948. BIOGRAPHIES OF A UTHORS I Gede P asek Suta W ijaya recei v ed the B.Eng. de grees in Electrical Engineering from Gadjah Mada Uni v e rsity in 1997, M.Eng. de grees in Computer Informatics System from Gadjah Mada Uni v ersity in 2001, and Doctor of Engineering de grees in Computer Science from K umamoto uni- v ersity , Japan in 2010. During 1998-1999 he w ork ed in T o yota Astra Motor Compan y in Indonesia as Planning Production Control, and from 1999-2000, ne xt, he w ork ed as lecturer assistance in Y o- gyakarta National T echnology Colle ge in Indonesia, and since 2000 up today , he has been full time lecturer and stays in Expert Systems Laboratory in Informatics Engineering Department, Mataram Uni v ersity , Indonesia. His research interests are pattern recognition, artificial intelligence, and im- age processing application on computer vision. K eiichi Uchimura recei v ed the B.Eng. and M.Eng. de grees from K umamoto Uni v ersity , K u- mamoto, Japan, in 1975 and 1977, respecti v ely , and the Ph.D. de gree from T ohoku Uni v ersity , Miyagi, Japan, in 1987. He is currently a Professor with the Graduate School of Science and T ech- nology , K umamoto Uni v ersity . He is eng aged in research on intelligent transportation systems, and computer vision. From 1992 to 1993, he w as a V isiting Researcher at McMaster Uni v ersity , Hamilton, ON, Canada. His research interests are computer vision and optimization problems in the Intelligence T ransport System. Gou K outaki ecei v ed the B.Eng., M.Eng., and Ph.D.de gree from K umamoto Uni v ersity , K u- mamoto, Japan, in 2002, 2004, and 2007, respecti v ely . From 2007 to 2010, he w as with Production Engineering Research Laboratory , Hitachi Ltd. He is currently an Assistant Professor with the Graduate School of Science and T echnology , K umamoto Uni v ersity . He is eng aged in research on image processing and pattern recognition of the Intelligence T ransport System. T r af fic Light Signal P ar ameter s Optimization Using Modification ... (I Gede P asek Suta W ijaya) Evaluation Warning : The document was created with Spire.PDF for Python.