A hybrid predictive control formulation based on evolutionary multi-objective optimization to optimize real-time operations of public transport systems is presented. The state space model includes bus position, expected load and arrival time at stops. The system is based on discrete events, and the possible operator control actions are: holding vehicles at stations and skipping some stations. The controller (operator) pursues the minimization of a dynamic objective function to generate better operational decisions under uncertain demand at bus stops. In this work, a multi-objective approach is conducted to include different goals in the optimization process that could be opposite. In this case, the optimization was defined in terms of two objectives: waiting time minimization on one side, and impact of the strategies on the other. A genetic algorithm method is proposed to solve the multi-objective dynamic problem. From the conducted experiments considering a single bus line corridor, we found that the two objectives are opposite but with a certain degree of overlapping, in the sense that in all cases both objectives significantly improve the level of service with respect to the open-loop scenario by regularizing the headways. On average, the observed trade-off validates the proposed multi-objective methodology for the studied system, allowing dynamically finding the pseudo-optimal Pareto front and making real-time decisions based on different optimization criteria reflected in the proposed objective function compounds.@en

In this article, a hybrid predictive control (HPC) strategy is formulated for the real-time optimisation of a public transport system operation run using buses. For this problem, the hybrid predictive controller corresponds to the bus dispatcher, who dynamically provides the optimal control actions to the bus system to minimise users' total travel time (on-vehicle ride time plus waiting time at stops). The HPC framework includes a dynamic objective function and a predictive model of the bus system, written in discrete time, where events are triggered when a bus arrives at a bus stop. Upon these events, the HPC controller makes decisions based on two well-known real-time transit control actions, holding and expressing. Additionally, the uncertain passenger demand is included in the model as a disturbance and then predicted based on both offline and online information of passenger behaviour. The resulting optimisation problem of the HPC strategy at every event is Np-hard and needs an efficient algorithm to solve it in terms of computation time and accuracy. We chose an ad hoc implementation of a Genetic Algorithm that permits the proper management of the trade-off between these two aspects. For real-time implementation, the design of this HPC strategy considers newly available transport technology such as the availability of automatic passenger counters (APCs) and automatic vehicle location (AVL) devices. Illustrative simulations at 2, 5 and 10 steps ahead are conducted, and promising results showing the advantages of the real-time control schemes are reported and discussed.@en

This paper presents a hybrid adaptive predictive control approach to incorporate future information regarding unknown demand and expected traffic conditions, in the context of a dynamic pickup and delivery problem with fixed fleet size. As the routing problem is dynamic, several stochastic effects have to be considered within the analytical expression of the dispatcher assignment decision objective function. This paper is focused on two issues: one is the extra cost associated with potential rerouting arising from unknown requests in the future, and the other is the potential uncertainty in travel time coming from non-recurrent traffic congestion from unexpected incidents. These effects are incorporated explicitly in the objective function of the hybrid predictive controller. In fact, the proposed predictive control strategy is based on a multivariable model that includes both discrete/integer and continuous variables. The vehicle load and the sequence of stops correspond to the discrete/integer variable, adding the vehicle position as an indicator of the traffic congestion conditions. The strategy is analyzed under two scenarios. The first one considers a predictable congestion obtained using historical data (off-line method) requiring a predictive model of velocities distributed over zones. The second scenario that accepts unpredictable congestion events generates a more complex problem that is managed by using both fault detection and isolation and fuzzy fault-tolerant control approaches. Results validating these approaches are presented through a simulated numerical example.@en

In this paper, we develop a family of solution algorithms based upon computational intelligence for solving the dynamic multi-vehicle pick-up and delivery problem formulated under a hybrid predictive adaptive control scheme. The scheme considers future demand and prediction of expected waiting and travel times experienced by customers. In addition, this work includes an analytical formulation of the proposed prediction models that allow us to search over a reduced feasible space. Predictive models consider relevant state space variables as vehicle load and departure time at stops. A generic expression of the system cost function is used to measure the benefits in dispatching decisions of the proposed scheme when solving for more than two-step ahead under unknown demand. The demand prediction is based on a systematic fuzzy clustering methodology, resulting in appropriate call probabilities for uncertain future. As the dynamic multi-vehicle routing problem considered is NP-hard, we propose the use of genetic algorithms (GA) that provide near-optimal solutions for the three, two and one-step ahead problems. Promising results in terms of computation time and accuracy are presented through a simulated numerical example that includes the analysis of the proposed fuzzy clustering, and the comparison of myopic and new predictive approaches solved with GA.@en

This paper presents a hybrid adaptive predictive control approach that includes future information in realtime routing decisions in the context of a dynamic pickup and delivery problem (DPDP). We recognize in this research that when the problem is dynamic, an additional stochastic effect has to be considered within the analytical expression of the objective function for vehicle scheduling and routing, which is the extra cost associated with potential rerouting arising from unknown requests in the future. The major contributions of this paper are: first, the development of a formal adaptive predictive control framework to model the DPDP, and second, the development and coding of an ad hoc particle swarm optimization (PSO) algorithm to efficiently solve it. Predictive state-space formulations are written on the relevant variables (vehicle load and departure time at stops) for the DPDP. Next, an objective function is stated to solve the real-time system when predicting one and two steps ahead in time. A problem-specific PSO algorithm is proposed and coded according to the dynamic formulation. Then, the PSO method is used to validate this approach through a simulated numerical example.@en

In this paper, a dynamic vehicle routing problem (DVRP) is solved based on hybrid predictive control strategy with an objective function that includes two dimensions: user and operator costs. To handle some undesired assignments for the users, a new objective function is designed, able to carry out the fact that some users can become particularly annoyed if their service is postponed. Genetic algorithms are proposed for efficiently solving the DVRP. Fuzzy clustering is applied for computing trip patterns from historical data under more realistic scenarios. An illustrative experiment through simulation of the process is presented to show the potential benefits (mainly for users) of the new design.@en

Hybrid Predictive Control for Dynamic Transport Problems develops methods for the design of predictive control strategies for nonlinear-dynamic hybrid discrete-/continuous-variable systems. The methodology is designed for real-time applications, particularly the study of dynamic transport systems. Operational and service policies are considered, as well as cost reduction. The control structure is based on a sound definition of the key variables and their evolution. A flexible objective function able to capture the predictive behaviour of the system variables is described. Coupled with efficient algorithms, mainly drawn from the area of computational intelligence, this is shown to optimize performance indices for real-time applications. The framework of the proposed predictive control methodology is generic and, being able to solve nonlinear mixed-integer optimization problems dynamically, is readily extendable to other industrial processes. The main topics of this book are: ●hybrid predictive control (HPC) design based on evolutionary multiobjective optimization (EMO); ●HPC based on EMO for dial-a-ride systems; and ●HPC based on EMO for operational decisions in public transport systems. Hybrid Predictive Control for Dynamic Transport Problems is a comprehensive analysis of HPC and its application to dynamic transport systems. Introductory material on evolutionary algorithms is presented in summary in an appendix. The text will be of interest to control and transport engineers working on the operational optimization of transport systems and to academic researchers working with hybrid systems. The potential applications of the generic methods presented here in other process fields will appeal to a wider group of researchers, scientists and graduate students working in other control-related disciplines.@en