Decentralized Traffic Signal Control for Grid Traffic ...

6
2019 6 th IEEE International Conference on Engineering Technologies and Applied Sciences (ICETAS) 978-1-7281-4082-7 /19/$31.00 ©2019 IEEE Decentralized Traffic Signal Control for Grid Traffic Network using Genetic Algorithm Min Keng Tan 1 , Helen Sin Ee Chuo 1 , Kiam Beng Yeo 2 , Renee Ka Yin Chin 1 , Sha Huang 3 , Kenneth Tze Kin Teo 1,* 1 Modelling, Simulation & Computing Laboratory | Artificial Intelligence Research Unit Faculty of Engineering, Universiti Malaysia Sabah Kota Kinabalu, Malaysia Email: [email protected], [email protected], [email protected] 2 Faculty of Medicine and Health Sciences Universiti Malaysia Sabah Kota Kinabalu, Malaysia 3 School of Rail Transportation Wuyi University Jiangmen, China Abstract This work aims to explore the potential to minimize traffic congestion using a non-deterministic algorithm. Conventionally, the deterministic algorithm such as fuzzy logic was proposed as the computational algorithm to compute the optimum traffic signal timing for minimizing vehicles in queue and travel delay. However, it is very difficult to define the suitable number of fuzzy rules that are able to cover the all possibilities of traffic flow changes since the natural traffic flow behavior is dynamic. Besides, the inherent deterministic behavior limits the algorithm to explore the solution space in searching for the optimum traffic solution. In other words, the deterministic algorithm will not provide other solution with the same input. Therefore, genetic algorithm, a non-deterministic algorithm, is proposed to optimize the traffic signalization. A benchmarked 3×3 grid traffic network is developed as the testbed to examine the robustness of the proposed GA. Each intersection is integrated with a GA based signal controller or known as agent to form a multi-agent system. Each agent has the autonomy in controlling their own traffic intersection and they will share their local traffic information to their downstream intersections. The performance of the proposed GA is compared with the conventional fuzzy logic. The simulation results show the proposed GA improves the performance about 6.6 % in minimizing vehicles in queue and travel delay as compared to the conventional fuzzy logic. Keywords decentralized traffic signal control, multi-agent system, grid traffic network, fuzzy logic, genetic algorithm I. INTRODUCTION Reducing traffic congestion is one of the key challenges to all the major cities for improving its socio-economy and society welfare [1]. Due to the space constraint in urban areas which limits the local authorities to build more or expand the existing infrastructures, optimizing the traffic signal timing is proven as cost-effective way to relieve congestion, especially for those cities that facing low public transport modal share [2,3]. Nowadays, many adaptive traffic signal controllers are reported in the literature to auto regulate the traffic signal timing based on the instant traffic demand. The deterministic algorithm, such as split, cycle, offset optimization technique (SCOOT) [4], Sydney coordinated adaptive traffic system (SCATS) [5] and fuzzy logic [6,7], is the most common reported computational algorithm. Although the deterministic algorithm is guaranteed to produce precise solutions [8,9], the deterministic behavior limits the algorithm to explore the solution space for seeking the optimum solution. For example, fuzzy logic based traffic signal controller will compute the traffic solution based on the predefined input and output membership functions as well as the fuzzy rules [10,11]. The fuzzy logic will always produce the same solution by given particular inputs. In order words, the algorithm determines the solution based on the predefined computational structure/process. It won’t be able to explore the solution space for seeking for other solutions which might be the optimum solution. This is because it is very difficult to developed well-defined rules that are able to describe all the traffic scenarios since the traffic behavior is dynamic in nature. Hence, this paper aims to explore the potential of enhancing the performance of conventional adaptive traffic signal control system using a non-deterministic computational algorithm. Genetic algorithm (GA) is proposed as the algorithm to optimize the traffic signal timing via the evolutionary process. Robustness of the proposed algorithm is tested using a benchmarked 3×3 grid traffic network model and the performance is compared with the conventional fuzzy logic. In this work, it is assumed that all the intersections are integrated with identical traffic signal controller or known as agent. Each agent has their own autonomy in controlling the local traffic flow by alternating the traffic signalization. In order to enhance the collaboration between the agents, they will share their local traffic information to their downstream intersections. This paper is organized as follow: Section II presents the modeling of grid traffic network which will be used as the simulation platform to examine the performance of decentralized multi-agent traffic signal control systems; Section III explains the development of proposed genetic algorithm based traffic signal controller; Section IV describes the conventional fuzzy logic based traffic signal controller; Section V discusses the performances of the both controllers in minimizing vehicles in queue and travel delay; and Section V concludes the findings. * Corresponding author This work is supported by FRGS grant, grant no. FRG0405-TK-2/2014 and SGI grant, grant no. SGI0089-2019. Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Transcript of Decentralized Traffic Signal Control for Grid Traffic ...

Page 1: Decentralized Traffic Signal Control for Grid Traffic ...

2019 6th IEEE International Conference on Engineering Technologies and Applied Sciences (ICETAS)

978-1-7281-4082-7 /19/$31.00 ©2019 IEEE

Decentralized Traffic Signal Control for

Grid Traffic Network using Genetic Algorithm

Min Keng Tan1, Helen Sin Ee Chuo1, Kiam Beng Yeo2, Renee Ka Yin Chin1, Sha Huang3, Kenneth Tze Kin Teo1,*

1Modelling, Simulation & Computing Laboratory | Artificial Intelligence Research Unit

Faculty of Engineering, Universiti Malaysia Sabah

Kota Kinabalu, Malaysia

Email: [email protected], [email protected], [email protected]

2Faculty of Medicine and Health Sciences

Universiti Malaysia Sabah

Kota Kinabalu, Malaysia

3School of Rail Transportation

Wuyi University

Jiangmen, China

Abstract — This work aims to explore the potential to

minimize traffic congestion using a non-deterministic

algorithm. Conventionally, the deterministic algorithm such as

fuzzy logic was proposed as the computational algorithm to

compute the optimum traffic signal timing for minimizing

vehicles in queue and travel delay. However, it is very difficult

to define the suitable number of fuzzy rules that are able to

cover the all possibilities of traffic flow changes since the

natural traffic flow behavior is dynamic. Besides, the inherent

deterministic behavior limits the algorithm to explore the

solution space in searching for the optimum traffic solution. In

other words, the deterministic algorithm will not provide other

solution with the same input. Therefore, genetic algorithm, a

non-deterministic algorithm, is proposed to optimize the traffic

signalization. A benchmarked 3×3 grid traffic network is

developed as the testbed to examine the robustness of the

proposed GA. Each intersection is integrated with a GA based

signal controller or known as agent to form a multi-agent

system. Each agent has the autonomy in controlling their own

traffic intersection and they will share their local traffic

information to their downstream intersections. The

performance of the proposed GA is compared with the

conventional fuzzy logic. The simulation results show the

proposed GA improves the performance about 6.6 % in

minimizing vehicles in queue and travel delay as compared to

the conventional fuzzy logic.

Keywords — decentralized traffic signal control, multi-agent

system, grid traffic network, fuzzy logic, genetic algorithm

I. INTRODUCTION

Reducing traffic congestion is one of the key challenges to all the major cities for improving its socio-economy and society welfare [1]. Due to the space constraint in urban areas which limits the local authorities to build more or expand the existing infrastructures, optimizing the traffic signal timing is proven as cost-effective way to relieve congestion, especially for those cities that facing low public transport modal share [2,3].

Nowadays, many adaptive traffic signal controllers are reported in the literature to auto regulate the traffic signal timing based on the instant traffic demand. The deterministic algorithm, such as split, cycle, offset optimization technique

(SCOOT) [4], Sydney coordinated adaptive traffic system (SCATS) [5] and fuzzy logic [6,7], is the most common reported computational algorithm. Although the deterministic algorithm is guaranteed to produce precise solutions [8,9], the deterministic behavior limits the algorithm to explore the solution space for seeking the optimum solution. For example, fuzzy logic based traffic signal controller will compute the traffic solution based on the predefined input and output membership functions as well as the fuzzy rules [10,11]. The fuzzy logic will always produce the same solution by given particular inputs. In order words, the algorithm determines the solution based on the predefined computational structure/process. It won’t be able to explore the solution space for seeking for other solutions which might be the optimum solution. This is because it is very difficult to developed well-defined rules that are able to describe all the traffic scenarios since the traffic behavior is dynamic in nature.

Hence, this paper aims to explore the potential of enhancing the performance of conventional adaptive traffic signal control system using a non-deterministic computational algorithm. Genetic algorithm (GA) is proposed as the algorithm to optimize the traffic signal timing via the evolutionary process. Robustness of the proposed algorithm is tested using a benchmarked 3×3 grid traffic network model and the performance is compared with the conventional fuzzy logic. In this work, it is assumed that all the intersections are integrated with identical traffic signal controller or known as agent. Each agent has their own autonomy in controlling the local traffic flow by alternating the traffic signalization. In order to enhance the collaboration between the agents, they will share their local traffic information to their downstream intersections.

This paper is organized as follow: Section II presents the modeling of grid traffic network which will be used as the simulation platform to examine the performance of decentralized multi-agent traffic signal control systems; Section III explains the development of proposed genetic algorithm based traffic signal controller; Section IV describes the conventional fuzzy logic based traffic signal controller; Section V discusses the performances of the both controllers in minimizing vehicles in queue and travel delay; and Section V concludes the findings.

* Corresponding author

This work is supported by FRGS grant, grant no. FRG0405-TK-2/2014 and SGI grant, grant no. SGI0089-2019.

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Page 2: Decentralized Traffic Signal Control for Grid Traffic ...

II. MODELING OF GRID TRAFFIC NETWORK

The benchmarked 3×3 traffic network model [12] is used as the simulation platform to examine the robustness of the proposed GA and conventional fuzzy logic based traffic signal controllers. Fig. 1 illustrates the topology of the benchmarked grid traffic network. As shown in the figure, the grid traffic network consists of nine identical at-grade four-legged intersections which are connected in three rows with each row consists of three intersections. Each legged or known as approach of the intersection comprises two through lanes, and one exclusive left- and right-turn lane.

In order to ensure the model is replicate Malaysian signalized intersection, all its geometric parameters, such as road width, turning radius and designed speed, are taken from the manual given by Public Works Department Malaysia (JKR) [13,14]. The intersection geometric design will affect the traffic outflow rate and the capacity of the intersection [15,16].

Normal distribution is used to generate the probability of inflow and outflow vehicles at each intersection, as presented in (1).

(1)

where is probability density function (PDF); is vehicle

flow rate; is average vehicle flow rate and is standard deviation of vehicle flow rate respectively. For the inflow

traffic model, the is defined as the average traffic demand

of each approach, whereas the is defined as saturated flow

rate of each approach for the outflow traffic model. The is

assumed to be 10 % of the .

The saturation flow rate is defined as the total number of vehicles can traverse without interruption and can be represented as (2).

S = N × So × fg × fa (2)

where S is saturation flow rate; N is number of lanes; So is base saturation flow rate; fg is adjustment factor for approach grade; and fa is adjustment factor for area type. Fig. 2 illustrates the transition of traffic phase per cycle.

In this work, the traffic signal of each intersection is controlled by an agent which has integrated with AI to calculate the traffic signal timing. Each agent has the autonomy in controlling their local intersection based on the local traffic flow information. Besides, agents will share their local traffic information to their downstream intersections for enhancing the collaboration between intersections. Fig. 3 illustrates the schematic diagram of decentralized multi-agent traffic signal control system.

Fig. 2. Transition of traffic phase per cycle [17].

Fig. 1. Topology of grid traffic network.

Fig. 3. Schematic diagram of decentralized multi-agent system.

Phase 1 Phase 2

Phase 3 Phase 4

Sim

ula

tio

n E

nvir

on

men

t

Intersection 1

Local Agent

Intersection 2

Local Agent

Intersection 3

Local Agent

Intersection 9

Local Agent

Traffic State

Signal

Traffic State

Signal

Traffic State

Signal

Traffic State

Signal

Traffic

State

.

.

.

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Page 3: Decentralized Traffic Signal Control for Grid Traffic ...

III. GENETIC ALGORITHM BASED SIGNAL CONTROLLER

The proposed GA exploits probabilistic search method to optimize the traffic signal based on the evolutionary theory: survival of the fittest [18,19]. Basically, the GA comprises three main mechanisms, namely evaluation, crossover and mutation. The flowchart of the GA based traffic signal controller is illustrated as Fig. 4. The proposed GA is used to calculate the near optimum traffic signal timings for each traffic phase in order to minimize the traffic queue and travel delay. There are 50 chromosomes (potential solutions) in the population of the proposed GA. Each of the chromosome is encoded with the information of cycle time which is ranged from 45 to 120 seconds [20].

The fitness of each chromosome is evaluated using the fitness function stated in (3). Based on the (3), the fitness value, fv is calculated based on the reciprocal of the average travel delay, td., where the average delay can be defined as (4). As such, the chromosome that leads to smallest td will receive highest fv.

fv = 1 / avg (td) (3)

avg (td) = [ ∑ ( td i ∙ vehin

i ) ] / [ ∑ ( vehin i ) ] (4)

where avg (td) is average time delay, td i is time delay of lane

group i; and vehin i is total inflow vehicles of lane group i.

During the reproduction process, the chromosomes with higher f is preferentially to be selected as the parents. Rank selection technique is used to avoid the fittest chromosome to be dominated in the selection process [21,22]. With the predefined crossover probability, Px, both selected parents will proceed with crossover process based on the (5) and (6). In order to keep diversity in the population, the newly generated offspring, with predefined mutation probability, Pm, might be mutated [23,24].

Offspring 1 = (α) (Parent 1) + (1 – α) (Parent 2) (5)

Offspring 2 = (1 – α) (Parent 1) + (α) (Parent 2) (6)

where α is the crossover factor. Table I summarizes the parameters of GA based traffic signal controller.

TABLE I. PARAMETERS OF GENETIC ALGORITHM

Parameters Symbol Typical Value

Number of population - 50

Stopping criteria - 100 iterations

Crossover probability Px 0.90

Mutation probability Pm 0.01

Solution space (cycle time) tC 45 ≤ tC ≤ 120 seconds

Fig. 4. Flowchart of genetic algorithm based traffic signal controller.

Start

Initialize the potential solutions

(chromosomes)

Evaluate the fitness

of each chromosome

Select 2 parents

Finish evaluate all chromosomes?

Determine optimum

solution

End

Yes

Yes

No

No

Crossover

Reach stopping criterion?

Generated rand.

num. ≤ Px?

Yes

No

Yes

No

Mutate

Generated rand.

num. ≤ Pm?

Yes

No

No. of offspring = population size?

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Page 4: Decentralized Traffic Signal Control for Grid Traffic ...

IV. FUZZY LOGIC BASED SIGNAL CONTROLLER

The performance of the proposed GA based traffic signal controller is compared with the conventional adaptive signal controller – fuzzy logic. This section discusses the development of the conventional fuzzy logic based traffic signal controller. The fuzzy logic is developed to calculate the optimum green time, tG, for each traffic phase based on the number of vehicles in queue, vehq and the queue discharge rate, ∆vehq. Fig. 5 illustrates the diagram of fuzzy logic based traffic signal control system. Generally, the operation of fuzzy logic can be divided into three operations, namely fuzzification, inference engine and defuzzification [25].

The fuzzification operation converts the inputs into linguistic fuzzy set using the input membership functions. As shown in Fig. 6(a) and (b), each input will have their own membership functions to convert the respective input into the linguistic input for the inference engine. In this work, each input comprises four membership functions. For example, the universe discourse of vehq is divided into four regions, namely Q1, Q2, Q3 and Q4.

The inference engine determines the logical decision in terms of linguistic output based on the predefined fuzzy rules. Since each input has four membership functions, the minimum number of rules required to cover all the functions is 4×4 = 16. Table II tabulates the predefined fuzzy rules for the signal controller. This work implements Mamdami type of fuzzy inference system (FIS), whereas the implication and aggregation techniques used are AND and OR operators respectively, as summarized in Table III.

The defuzzification converts the linguistic fuzzy sets into actual tG. In this work, the universe discourse of the tG is divided into 7 regions, namely Min, Shortest, Shorter, Medium, Long, Longest and Max, as shown in Fig. 6(c). As mentioned in previously, the maximum limit of cycle time is 120 seconds. Hence, each traffic phase in this work is limited to 120 seconds / 4 traffic phase = 30 seconds per phase.

TABLE II. FUZZY RULES FOR TRAFFIC SIGNAL CONTROLLER

vehq

Q1 Q2 Q3 Q4

∆ve

hq

Discharge Shortest Medium Shorter Long

SDischarge Min Medium Shortest Long

SCharge Min Longest Medium Max

Charge Shorter Long Shorter Longest

TABLE III. SUMMARY OF THE IMPLEMENTED TECHNIQUES

Technique

Fuzzy inference system (FIS) Mamdani

Implication Minimum [AND]

Aggregation Maximum [OR]

Defuzzification Centroid

Fig. 5. Diagram of fuzzy logic based traffic signal control system.

(a) Input membership functions for vehicles in queue, vehq

(b) Input membership functions for change in vehicles in queue, ∆vehq

(c) Output membership functions for green time, tG

Fig. 6. Membership functions for fuzzy logic based signal controller.

Fuzzy Rules

Fuzzification

Inference Engine

Defuzzification

∆vehq

tG vehq

Inputs Output

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Page 5: Decentralized Traffic Signal Control for Grid Traffic ...

V. RESULTS AND DISCUSSION

This section discusses the performances of the proposed GA and the conventional fuzzy logic. The robustness of the both traffic signal controllers is tested under oversaturated condition, where the traffic demand is 50.0 % higher than the nominal traffic. In this work, the behaviors of non-uniform headway for inflow and outflow vehicles as well as the driver’s perception-reaction delay which is in between 0 and 7 seconds are simulated. Besides, the propagation effect of traffic congestion is also considered in the simulation. Fig. 7 and 8 present the performances of the conventional fuzzy logic controller and the proposed GA controller respectively, whereas the Table IV summarizes the performance measurement of both controllers.

From the results, it can be observed that the fuzzy logic fails to minimize the queued vehicles. As the results, the average travel delay increases over time [Fig. 7(b)] as the queue is growing [Fig. 7(a)]. The performance of fuzzy logic is inadequate because it is difficult to develop a set of fuzzy rules that can cover all the possibilities of traffic scenario. Besides, the inherent deterministic computational characteristic of fuzzy logic constrains the algorithm to explore the solution space for seeking the optimum solution.

TABLE IV. PERFORMANCE MEASUREMENT OF BOTH CONTROLLERS

Measured Parameters Fuzzy Logic Genetic Algorithm

Sum of vehin veh 48,108 48,007

Average vehq veh 8.691 8.019

Average td s 58.905 55.045

Fuel wastage l 881.6 822.1

CO2 emission kg 2,018.0 1,881.9

In contrast to the conventional fuzzy logic, the proposed GA controller has better performance in minimizing the queue level because the GA is a non-deterministic computational algorithm which can self-explore the solution space for seeking the optimum traffic solution. As shown in Fig. 8, the GA controller is able to minimize the average queue length by distributing the average delay among the intersection more effectively. As such, the average delay experienced by the proposed GA is 6.6 % lesser than the conventional fuzzy logic [Table IV]. As the conclusion, the proposed GA has better performance because it is not constrained by the deterministic computation and is able to explore the solution space for seeking the optimum solution.

(a) Average queue

(b) Average delay

Fig. 7. Performances of the conventional fuzzy logic controller.

(a) Average queue

(b) Average delay

Fig. 8. Performances of the proposed genetic algorithm controller.

Fail to

control

Delay

increase

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.

Page 6: Decentralized Traffic Signal Control for Grid Traffic ...

VI. CONCLUSION

This paper has developed a non-deterministic algorithm to optimize traffic network signalization. The genetic algorithm (GA) is proposed as the computational algorithm to calculate the optimum traffic signal timing. The robustness of the proposed GA is tested using the benchmarked 3×3 grid traffic network model. In this work, the proposed GA is integrated into the traffic signal controller of each intersection or known as agent. Each agent has the autonomy in controlling their local intersection and they will share their local traffic information to the downstream intersections for enhancing the collaboration among the agents. The performance of the proposed GA is compared with the conventional fuzzy logic based traffic signal control system, which is a deterministic control system. Both algorithms are tested under saturated traffic flow condition and the simulation results show the GA is able to explore the solution space for seeking the optimum traffic solution without limited by the deterministic computational process. As such, the proposed GA has improved the performance by 6.6 % in reducing vehicles in queue and minimizing travel delay as compared to the conventional fuzzy logic.

ACKNOWLEDGMENT

The authors would like to acknowledge the Ministry of Education Malaysia (KPM) for supporting this research under Fundamental Research Grant Scheme (FRGS), grant no. FRG0405-TK-2/2014 and MyPhD scholarship under MyBrain15 program, as well as Universiti Malaysia Sabah (UMS) for supporting this research under Innovation Grant Scheme (SGI), grant no. SGI0089-2019.

REFERENCES

[1] M.K. Tan, H.S.E. Chuo, R.K.Y. Chin, K.B. Yeo, and K.T.K. Teo, “Optimization of traffic network signal timing using decentralized genetic algorithm,” IEEE 2nd International Conference on Automatic Control and Intelligent Systems, pp. 62-67, 2017, doi: 10.1109/I2CACIS.2017.8239034.

[3] F.G. Sander, L.C. Blancas, and R. Westra, Malaysia Economic Monitor: Transforming Urban Transport, Washington, DC: The World Bank, 2015, report no. 97393.

[4] M.K. Ekbatani, M. Papageorgiou, and V.L. Knoop, “Controller design for gating traffic control in presence of time-delay in urban road networks,” Transportation Research Part C: Emerging Technologies, vol. 59, pp. 308-322, 2015, doi: 10.1016/j.trc.2015.04.031.

[5] D. Zhao, Y. Dai, and Z. Zhang, “Computational intelligence in urban traiffc signal control: A survey. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), vol. 42, no. 4, pp. 485-494, 2012, doi: 10.1109/TSMCC.20112161577.

[6] H.S.E. Chuo, M.K. Tan, A.C.H. Chong, R.K.Y. Chin, and K.T.K. Teo, “Evolvable traffic signal control for intersection congestion alleviation with enhanced particle swarm optimisation,” IEEE 2nd International Conference on Automatic Control and Intelligent Systems, pp. 92-97, 2017, doi: 10.1109/I2CACIS.2017.8239039.

[7] K.T.K. Teo, H.H. Goh, B.L. Chua, S.K. Tang, and M.K. Tan, “Modelling and optimisation of stand alone power generation at rural area,” IEEE International Conference on Consumer Electronics-China, pp. 51-56, 2013, doi: 10.1109/ICCE-China.2013.6780868.

[8] C.S. Chin, P. Neelakantan, S.S. Yang, B.L. Chua, and K.T.K. Teo, “Effect of partially shaded conditions on photovoltaic array’s maximum power point tracking,” International Journal of Simulation – Systems, Science & Technology, vol. 12, no. 3, pp. 52-59, 2011, doi: 10.5013/IJSSST.a.12.03.08.

[9] K.T.K. Teo, K.B. Yeo, S.E. Tan, Z.W. Siew, and K.G. Lim, “Design and development of portable fuzzy logic based traffic optimizer,” IEEE International Conference on Consumer Electronics – China, pp. 7-12, 2013, doi: 10.1109/ICCE-China.2013.6780856.

[10] H. Homaei, S.R. Hejazi, and S.A.M. Dehghan, “A new traffic light controller using fuzzy logic for a full single junction involving emergency vehicle preemption,” Journal of Uncertain Systems, vol. 9, no. 1, pp. 49-61, 2015, ISSN: 1752-8917.

[11] L.A. Ajao, F.J. Ajao, M.A. Adegboye, and A.A. Ismail, “An embedded fuzzy logic based application for density traffic control system,” International Journal of Artificial Intelligence Research, vol. 2, no. 1, pp. 7-16, 2018, doi: 10.29099/ijair.v2i1.44.

[12] K. Jang, H. Kim and I.G. Jang, “Traffic signal optimization for oversaturated urban networks: Queue growth equalization,” IEEE Transactions on Intelligent Transporation Systems, vol. 16, no. 4, pp. 2121-2128, 2015, doi: 10.1109/TITS.2015.2398896.

[13] Public Works Department Malaysia, A Guide to the Design of Traffic Signals. Arahan Teknik (Jalan) 11/87. Kuala Lumpur: Ministry of Works Malaysia. 1987.

[14] K.T.K. Teo, K.B. Yeo, Y.K. Chin, H.S.E. Chuo, and M.K. Tan, “Agent-based traffic flow optimization at multiple signalized intersections,” 8th Asia Modelling Symposium, pp. 21-26, 2014, doi: 10.1109/AMS.2014.16.

[15] H.S.E. Chuo, M.K. Tan, B.L. Chua, R.K.Y. Chin, and K.T.K. Teo, “Computation of cell transmission model for congestion and recovery traffic flow,” IEEE International Conference on Consumer Electronics-Asia, pp. 1-4, 2016, doi: 10.1109/ICCE-Asia.2016.7804769.

[16] L. Angeline, M.Y. Choong, B.L. Chua, R.K.Y. Chin, and K.T.K. Teo, “A traffic cellular automaton model with optimised speed,” IEEE International Conference on Consumer Electronics-Asia, pp. 1-4, 2016, doi: 10.1109/ICCE-Asia.2016.7804790.

[17] M.K. Tan, H.S.E. Chuo, R.K.Y. Chin, K.B. Yeo, and K.T.K. Teo, “Optimization of urban traffic network signlization using genetic algorithm,” IEEE Conference on Open Systems, pp. 87-92, 2016, doi: 10.1109/ICOS.2016.7881994.

[18] K.T.K. Teo, R.K.Y. Chin, S.E. Tan, C.H. Lee, and K.G. Lim, “Exploration of genetic algorithm in network coding for wireless sensor networks,” International Journal of Simulation: Systems, Science & Technology, vol. 15, no. 6, pp. 83-89, 2014, doi: 10.5013/IJSSST.a.15.06.09.

[19] C.H. Lee, K.G. Lim, M.K. Tan, R.K.Y. Chin, and K.T.K. Teo, “A genetic algorithm for management of coding resources in VANET,” IEEE 2nd International Conference on Automatic Control and Intelligent Systems, pp. 80-85, 2017, doi: 10.1109/I2CACIS.2017.8239037.

[20] K.T.K. Teo, K.B. Yeo, Y.K. Chin, H.S.E. Chuo, and M.K. Tan, “Agent-based optimization for multiple signalized intersections using Q-learning,” International Journal of Simulation: Systems, Science & Technology, vol. 15, no. 6, pp. 90-96, 2014, doi: 10.5013/IJSSST.a.15.06.10.

[21] M.T. Asif, N. Mitrovic, J. Dauwels, and P. Jaillet, “Matrix and tensor based methods for missing data estimation in large traffic networks,” IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 7, pp. 1816-1825, 2016, doi: 10.1109/TITS.2015.2507259.

[22] K.T.K. Teo, R.K.Y. Chin, S.E. Tan, C.H. Lee, and K.G. Lim, “Performance analysis of enhanced genetic algorithm based network coding in wireless networks,” 8th Asia Modelling Symposium, pp. 213-218, 2014, doi: 10.1109/AMS.2014.49.

[24] C.H. Wong, Z.W. Siew, M.K. Tan, R.K.Y. Chin, and K.T.K. Teo, “Optimization of distributed and collaborative beamforming in wireless sensor networks,” 4th International Conference on Computational Intelligence, Communication Systems and Networks, pp. 84-89, 2012, doi: 10.1109/CICSyN.2012.26.

[25] C.S. Chin, Y.K. Chin, B.L. Chua, A. Kiring, and K.T.K. Teo, “Fuzzy logic based MPPT for PV array under partially shaded conditions,” International Conference on Advanced Computer Science Applications and Technologies, pp. 133-138, 2012, doi: 10.1109/ACSAT.2012.46.

[23] Ahmed, Syed Faiz, et al. "Model predictive control for upper limb rehabilitation robotic system under disturbed condition." AIP Conference Proceedings. Vol. 2129. No. 1. AIP Publishing LLC, 2019.

[2] Ahmed, Syed Faiz, et al. "Global optimal solution for active noise control problem." Indian Journal of Science and Technology 4.9 (2011): 1015-1020.

Authorized licensed use limited to: UNIVERSITY SABAH MALAYSIA. Downloaded on November 27,2020 at 16:50:20 UTC from IEEE Xplore. Restrictions apply.