With the advent of modular autonomous vehicles (MAVs), this paper presents a novel operational design for flex-route transit services to reduce operation costs of vehicles and improve the service quality of customers. The regime allows the simultaneous dispatch of a certain amount of MAVs from a bus terminal at a departure time. Each MAV is allowed to visit customers freely outside of checkpoints. Self-adaptive capacity and flexible service mode adapt time- and space-dependent demand characteristics. The presented operational design is formulated as a mixed-integer linear program that is NP-hard. A two-stage solution framework is developed to decompose the proposed mathematical programming cautiously. In the first stage, customized dynamic programming with valid cuts is designed to solve a bus scheduling problem efficiently. In the second stage, an effective and fast heuristic is proposed to solve a variant of the dial-a-ride problem and satisfy the technical requirements for developing on-line applications. Numerical examples and a case study show the effectiveness of the proposed design by comparing the flex-route transit services using traditional vehicles.

Improving flex-route transit services with modular autonomous vehicles

Lane changing behavior has a significant impact on traffic efficiency and may lead to traffic delays or even accidents. It is important to plan a safe and efficient lane-changing trajectory that coordinates with the surrounding environment. Most conventional lane-changing models need to establish and solve constrained optimization models during the whole process, while reinforcement learning can just take the current state as input and directly output actions to vehicles. This study develops a lane-changing model using the deep deterministic policy gradient method, which can simultaneously control the lateral and longitudinal motions of the vehicle. To optimize its performance, a reward function is properly designed by combining safety, efficiency, gap, headway, and comfort features. To avoid collisions, a safety modification model is developed to check and correct acceleration at every time step. The driving trajectory data of 1169 lane-changing scenarios extracted from the Next Generation Simulation (NGSIM) dataset are used to train and test the model. The proposed model can quickly converge in training phase. Testing results show it can complete safe and efficient lane changing in different lane-changing scenarios with both shorter time headway and lane-changing duration than human drivers. Compared with the conventional dynamic lane-changing trajectory planning model, our model can reduce collision risk. It is also evaluated in automated and nonautomated mixed traffic in SUMO. Simulation results show that the proposed model also has a positive effect on the average speed of overall traffic flow.

A Novel Dynamic Lane-Changing Trajectory Planning Model for Automated Vehicles Based on Reinforcement Learning

A cooperative merging speed control strategy of CAVs based on virtual platoon in on-ramp merging system

The full implementation of electric public transport is a key step for the transport industry to move toward electrification and achieve carbon neutrality. However, in the face of time-varying demands and high-quality service requirements, traditional transit systems are difficult to ascend as the preferred mode of travel due to the constraints of fixed vehicle capacity and multiline transfers. With the advent of modular vehicle technology, it is becoming more realistic to develop an entirely new transportation system based on electric modular vehicles (EMVs). This study proposes a novel intelligent platoon transit system (IPTS), and its overall concept and operating mode are elaborated at a strategic level. In particular, the electrical, modular, and autonomous platoon transit system should be designed to achieve adaptive adjustment of capacity and possible en route transfers, which significantly improves the convenience, flexibility, and economy of public transport. We also design three application scenarios with varying demands during multistage development to bridge the gap in traditional buses. The key issues and case applicability of the three scenarios are discussed.

https://www.mdpi.com/2032-6653/13/8/153/pdf?version=1660296057

Design of an Intelligent Platoon Transit System towards Transportation Electrification

Traffic congestion has caused trouble for people for many years. With the level of traffic increasing in previous years, local, state and national governments, along with researchers, are looking for a way to manage this problem. Ramp Metering is a method developed to reduce highway traffic by controlling the flow of vehicles. What started as a manual process has now technically advanced, with the rise of AI creating adaptive solutions. However, traditional AI methods require a lot of data and communication power to carry out this. With data protection rules and regulations on the rise, this data might be difficult to source. Therefore, adaptive models are complicated to train and, as a result, create. The proposal of Federated Reinforcement Learning (FRL) seems to be a solution to this problem in recent years. Combining reinforcement learning with federated learning, FRL can create an adaptive solution by communicating with many different client nodes while protecting the privacy of the data. This thesis aims to create an FRL design for the problem of adaptive/intelligent ramp metering control. Through research into the background and related works of FRL, a design is proposed. This design is implemented by using SUMO, a simulation environment, and TraCI, an interface built to interact with SUMO. The design consists of a Q-Learning method integrated into an FL model. The design is then evaluated in three different scenarios; low traffic load, high traffic load and mixed traffic load. It is compared to industry-popular methods of ramp metering such as fixed-timing and Q-Learning. From the results, it can be concluded that there are no improvements in the network by using FRL. This could be due to a configuration error which will be detailed in Chapter 5. However, FRL still poses an excellent research opportunity, and further work in this area is detailed in the final chapter.

Intelligent Ramp Metering Control using Federated Reinforcement Learning

Ramp metering that uses traffic signals to regulate vehicle flows from the on-ramps has been widely implemented to improve vehicle mobility of the freeway. Previous studies generally update signal timings in real-time based on predefined traffic measurements collected by point detectors, such as traffic volumes and occupancies. Comparing with point detectors, traffic cameras—which have been increasingly deployed on road networks—could cover larger areas and provide more detailed traffic information. In this work, we propose a deep reinforcement learning (DRL) method to explore the potential of traffic video data in improving the efficiency of ramp metering. Vehicle locations are extracted from the traffic video frames and are reformed as position matrices. The proposed method takes the preprocessed video data as inputs and learns the optimal control strategies directly from the high-dimensional inputs. A series of simulation experiments based on real-world traffic data are conducted to evaluate the proposed approach. The results demonstrate that, in comparison with a state-of-the-practice method, the proposed DRL method results in (1) lower travel times in the mainline, (2) shorter vehicle queues at the on-ramp, and (3) higher traffic flows downstream of the merging area. The results suggest that the proposed method is able to extract useful information from the video data for better ramp metering controls.

/pdf/a-deep-reinforcement-learning-approach-for-ramp-metering-1577jizajx.pdf

A Deep Reinforcement Learning Approach for Ramp Metering Based on Traffic Video Data

The Modular Autonomous Vehicle (MAV) systems allow a vehicle module to join onto and detach from other modules to dynamically adjust vehicle capacity. It potentially renders transit agencies more flexibility to deal with the temporal fluctuations of passenger demand. In this work, we propose a strategy for flexible MAV scheduling on transit routes to meet the time-varying passenger demand. The proposed strategy is formulated as a bi-objective optimization model considering both the utilization of vehicles and service quality. The model determines the scheduled departure times from the terminals, the length of MAV for each scheduled trip, and the assignment of modules to all scheduled trips, simultaneously. The e-constraint method is adopted to solve the developed model and the fuzzy satisfying approach is employed to select the best possible solution. We implement the proposed strategy in a real-world case study in comparison with a traditional strategy to demonstrate the effectiveness of the proposed strategy. The results show that the proposed strategy can remarkably improve the utilization of vehicles and also make passengers more convenient. Specifically, it leads to an 84.9% reduction in the total empty seat, as well as a 12.62% reduction in the total passenger waiting time.

Scheduling strategy for transit routes with modular autonomous vehicles

Ramp metering that uses traffic signals to regulate vehicle flows from the on-ramps has been widely implemented to improve vehicle mobility of the freeway. Previous studies generally update signal timings in real-time based on predefined traffic measures collected by point detectors, such as traffic volumes and occupancies. Comparing with point detectors, traffic cameras-which have been increasingly deployed on road networks-could cover larger areas and provide more detailed traffic information. In this work, we propose a deep reinforcement learning (DRL) method to explore the potential of traffic video data in improving the efficiency of ramp metering. The proposed method uses traffic video frames as inputs and learns the optimal control strategies directly from the high-dimensional visual inputs. A real-world case study demonstrates that, in comparison with a state-of-the-practice method, the proposed DRL method results in 1) lower travel times in the mainline, 2) shorter vehicle queues at the on-ramp, and 3) higher traffic flows downstream of the merging area. The results suggest that the proposed method is able to extract useful information from the video data for better ramp metering controls.

Bing Liu

Papers

A Deep Reinforcement Learning Approach for Ramp Metering Based on Traffic Video Data

Scheduling strategy for transit routes with modular autonomous vehicles

A Deep Reinforcement Learning Approach for Ramp Metering Based on Traffic Video Data