Linsen Chong

A Computationally Efficient Simulation-Based Optimization Algorithm for Large-Scale Urban Transportation Problems

This paper proposes a computationally efficient simulation-based optimization SO algorithm suitable to address large-scale generally constrained urban transportation problems. The algorithm is based on a novel metamodel formulation. We embed the metamodel within a derivative-free trust region algorithm and evaluate the performance of this SO approach considering tight computational budgets. We address a network-wide traffic signal control problem using a calibrated microscopic simulation model of evening peak period traffic of the full city of Lausanne, Switzerland, which consists of more than 600 links and 200 intersections. We control 99 signal phases of 17 intersections distributed throughout the entire network. This SO problem is a high-dimensional nonlinear constrained problem. It is considered large-scale and complex in the fields of derivative-free optimization, traffic signal optimization, and simulation-based optimization. We compare the performance of the proposed metamodel method to that of a traditional metamodel method and that of a widely used commercial signal control software. The proposed method systematically and efficiently identifies signal plans with improved average city-wide travel times.

Simulation of Driver Behavior with Agent-Based Back-Propagation Neural Network

Two microscopic simulation methods are compared for driver behavior: the Gazis–Herman–Rothery (GHR) car-following model and a proposed agent-based neural network model. To analyze individual driver characteristics, a back-propagation neural network is trained with car-following episodes from the data of one driver in the naturalistic driving database to establish action rules for a neural agent driver to follow under perceived traffic conditions during car-following episodes. The GHR car-following model is calibrated with the same data set, using a genetic algorithm. The car-following episodes are carefully extracted and selected for model calibration and training as well as validation of the calibration rules. Performances of the two models are compared, with the results showing that at less than 10-Hz data resolution the neural agent approach outperforms the GHR model significantly and captures individual driver behavior with 95% accuracy in driving trajectory.

An efficient simulation-based optimization algorithm for large-scale transportation problems

This paper applies a computationally efficient simulation-based optimization (SO) algorithm suitable for large-scale transportation problems. The algorithm is based on a metamodel approach. The metamodel combines information from a high-resolution yet inefficient microscopic urban traffic simulator with information from a scalable and tractable analytical macroscopic traffic model. We then embed the model within a derivative-free trust region algorithm. We evaluate its performance considering tight computational budgets. We illustrate the efficiency of this algorithm by addressing an urban traffic signal control problem for the full city of Lausanne, Switzerland. The problem consists of a nonlinear objective function with nonlinear constraints. The problem addressed is considered large-scale and complex both in the fields of derivative-free optimization and simulation-based optimization. We compare the performance of the method to a traditional metamodel method.

A Simulation-Based Optimization Algorithm for Dynamic Large-Scale Urban Transportation Problems

This paper addresses large-scale urban transportation optimization problems with time-dependent continuous decision variables, a stochastic simulation-based objective function, and general analytical differentiable constraints. We propose a metamodel approach to address, in a computationally efficient way, these large-scale dynamic simulation-based optimization problems. We formulate an analytical dynamic network model that is used as part of the metamodel. The network model formulation combines ideas from transient queueing theory and traffic flow theory. The model is formulated as a system of equations. The model complexity is linear in the number of road links and is independent of the link space capacities. This makes it a scalable model suitable for the analysis of large-scale problems. The proposed dynamic metamodel approach is used to address a time-dependent large-scale traffic signal control problem for the city of Lausanne. Its performance is compared to that of a stationary metamodel approach. ...

Agent-based evaluation of driver heterogeneous behavior during safety-critical events

Heterogeneous driver behavior during safety-critical events is more complicated than normal driving situations and is difficult to capture by statistical models. This paper applies an agent-based reinforcement learning method to represent heterogeneous driving behavior for different drivers during safety-critical events. The naturalistic driving data of different drivers during safety-critical events are used in agent training. As an output of the Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) training technique, behavior rules are embedded in different agents to represent heterogeneous actions between drivers. The results show that the NFACRL is able to simulate naturalistic driver behavior and present heterogeneity.

Combined car-following and unsafe event trajectory simulation using agent based modeling techniques

This paper presents a research effort aimed at modeling normal and safety-critical driving behavior in traffic under naturalistic driving data using agent based modeling techniques. Neuro-fuzzy reinforcement learning was used to train the agents. The developed agents were implemented in the VISSIM simulation platform and were evaluated by comparing the behavior of vehicles with and without agent behavior activation. The results showed very close resemblance of the behavior of agents to driver data.

Determination and optimization of reinforcement learning parameters for driver actions in traffic

An agent-based, artificial intelligence technique known as reinforcement learning has been used to capture vehicle behavior and simulate drivers actions in traffic, especially during emergency situations. This paper discusses the training parameters and their influence on agent simulation performance. A type of agent training technique called Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) is used to test the training parameters with an objective of improving simulation performance. A systematic parameter determination and optimization methodology is provided.

A revised reinforcement learning algorithm to model complicated vehicle continuous actions in traffic

An agent-based multi-layer reinforcement learning (RL) framework for naturalistic driving behavior simulation in traffic is introduced. Each agent is a replication of an individual driver. Each agent is implemented by applying artificial intelligence concepts, including: fuzzy logic, neural networks, and reinforcement learning algorithms. A revised Neuro-Fuzzy Actor Critic Reinforcement Learning (NFACRL) is proposed to simulate vehicle actions during safety-critical events when the traffic state is complicated. The revised NFACRL algorithm can handle state dimension problems and continuous vehicle actions.

Neuro-Fuzzy Actor Critic Reinforcement Learning for determination of optimal timing plans

The purpose of timing plan optimization is to decrease delay and increase the overall performance of transportation network. This paper presents an agent-based reinforcement learning framework to train optimization agents to take appropriate actions according to perceived traffic states. Neuro-Fuzzy Actor-Critic Reinforcement Learning (NFACRL) method is applied in isolated intersection control. The control agent gets knowledge of traffic states after the learning process and determines the optimal phase durations required to minimize vehicle delay at a given intersection.