Stefano Di Cairano

Stochastic MPC With Learning for Driver-Predictive Vehicle Control and its Application to HEV Energy Management

This paper develops an approach for driver-aware vehicle control based on stochastic model predictive control with learning (SMPCL). The framework combines the on-board learning of a Markov chain that represents the driver behavior, a scenario-based approach for stochastic optimization, and quadratic programming. By using quadratic programming, SMPCL can handle, in general, larger state dimension models than stochastic dynamic programming, and can reconfigure in real-time for accommodating changes in driver behavior. The SMPCL approach is demonstrated in the energy management of a series hybrid electrical vehicle, aimed at improving fuel efficiency while enforcing constraints on battery state of charge and power. The SMPCL controller allocates the power from the battery and the engine to meet the driver power request. A Markov chain that models the power request dynamics is learned in real-time to improve the prediction capabilities of model predictive control (MPC). Because of exploiting the learned pattern of the driver behavior, the proposed approach outperforms conventional model predictive control and shows performance close to MPC with full knowledge of future driver power request in standard and real-world driving cycles.

Power Smoothing Energy Management and Its Application to a Series Hybrid Powertrain

Energy management strategies in hybrid electric vehicles determine how much energy is produced/stored/used in each powertrain component. We propose an approach for energy management applied to a series hybrid electric vehicle that aims at improving the powertrain efficiency rather than the total fuel consumption. Since in the series configuration the engine is mechanically decoupled from the traction wheels, for a given power request the steady-state engine operating point is chosen to maximize the efficiency. A control algorithm regulates the transitions between different operating points by using the battery to smoothen the engine transients, thereby improving efficiency. Because of the constrained nature of the transient-smoothing problem, we implement the control algorithm by model predictive control. The control strategy feedback law is synthesized and integrated with the powertrain control software in the engine control unit. Simulations of the urban dynamometer driving schedule (UDDS) and US06 cycles using a complete vehicle system model and experimental tests of the UDDS cycle show improved fuel economy with respect to baseline strategies.

Hybrid Modeling, Identification, and Predictive Control: An Application to Hybrid Electric Vehicle Energy Management

Rising fuel prices and tightening emission regulations have resulted in an increasing need for advanced powertrain systems and systematic model-based control approaches. Along these lines, this paper illustrates the use of hybrid modeling and model predictive control for a vehicle equipped with an advanced hybrid powertrain. Starting from an existing high fidelity nonlinear simulation model based on experimental data, the hybrid dynamical model is developed through the use of linear and piecewise affine identification methods. Based on the resulting hybrid dynamical model, a hybrid MPC controller is tuned and its effectiveness is demonstrated through closed-loop simulations with the high-fidelity nonlinear model.

Reference and command governors: A tutorial on their theory and automotive applications

This paper provides a tutorial overview of reference governors and command governors, which are add-on control schemes for reference supervision and constraint enforcement in closed-loop feedback control systems. The main approaches to the development of such schemes for linear and nonlinear systems are described. The treatment of unmeasured disturbances and parametric uncertainties is addressed. Generalizations to extended command governors, feedforward reference governors, reduced order reference governors, parameter governors, networked reference governors, decentralized reference governors, and virtual state governors are summarized. Examples of applications of these techniques to automotive systems are given. A comprehensive list of references is included. Comments comparing reference and command governor approaches with Model Predictive Control and input shaping, and on future directions in reference and command governor research are included.

Projection-free parallel quadratic programming for linear model predictive control

A key component in enabling the application of model predictive control (MPC) in fields such as automotive, aerospace, and factory automation is the availability of low-complexity fast optimisation algorithms to solve the MPC finite horizon optimal control problem in architectures with reduced computational capabilities. In this paper, we introduce a projection-free iterative optimisation algorithm and discuss its application to linear MPC. The algorithm, originally developed by Brand for non-negative quadratic programs, is based on a multiplicative update rule and it is shown to converge to a fixed point which is the optimum. An acceleration technique based on a projection-free line search is also introduced, to speed-up the convergence to the optimum. The algorithm is applied to MPC through the dual of the quadratic program (QP) formulated from the MPC finite time optimal control problem. We discuss how termination conditions with guaranteed degree of suboptimality can be enforced, and how the algorithm performance can be optimised by pre-computing the matrices in a parametric form. We show computational results of the algorithm in three common case studies and we compare such results with the results obtained by other available free and commercial QP solvers.

Optimal control of discrete hybrid stochastic automata

This paper focuses on hybrid systems whose discrete state transitions depend on both deterministic and stochastic events. For such systems, after introducing a suitable hybrid model called Discrete Hybrid Stochastic Automaton (DHSA), different finite-time optimal control approaches are examined: (1) Stochastic Hybrid Optimal Control (SHOC), that “optimistically” determines the trajectory providing the best trade off between the tracking performance and the probability that stochastic events realize as expected, under specified chance constraints; (2) Robust Hybrid Optimal Control (RHOC) that, in addition, less optimistically, ensures that the system remains within a specified safety region for all possible realizations of stochastic events. Sufficient conditions for the asymptotic convergence of the state vector are given for receding-horizon implementations of the above schemes. The proposed approaches are exemplified on a simple benchmark problem in production system management.

A predictive control solution for driveline oscillations damping

This paper deals with the problem of damping driveline oscillations, which is crucial to improving driveability and passenger comfort. Recently, this problem has received an increased interest due to the introduction in several production vehicles of the dual-clutch powershift automatic transmission with dry clutches. This type of transmission improves fuel economy, but it results in a challenging control problem, due to driveline oscillations. These oscillations, also called shuffles, occur during gear-shift, while traversing backlash or when tip-in and tip-out maneuvers are performed. The first contribution of this paper is the derivation of an accurate piecewise affine drivetrain model with three inertias. The second contribution is concerned with the design of a horizon-1 predictive controller based on flexible Lyapunov functions. Several simulations based on realistic scenarios show that the proposed control scheme can handle both the performance and physical constraints, and the strict limitations on the computational complexity.

Cloud-computing based velocity profile generation for minimum fuel consumption: A dynamic programming based solution

This paper proposes a new framework to minimize the fuel consumed in a conventional vehicle over a given driving route by finding the optimal velocity profile. The optimization problem is solved in a remote cloud computing environment and assumes the vehicle route to be known a priori. A spatial domain dynamic programming optimization algorithm is used in this study to find the optimal velocity profile. The cloud-computing environment integrates information from GIS, road speed limits, into a vehicle simulator equipped with a fuel consumption model to predict fuel use along the desired route. The resulting global optimal velocity profile is sent back to the driver for velocity advisory.

Model predictive control for spacecraft rendezvous and docking with a rotating/tumbling platform and for debris avoidance

A Model Predictive Control (MPC) approach is developed for spacecraft rendezvous and docking to a rotating/tumbling platform and for debris avoidance maneuvers. With this approach, the constraints on thrust, approach velocity and spacecraft positioning within the Line-of-Sight cone from the docking port are systematically treated. The trajectories are simulated and time-to-dock and fuel consumption are evaluated as cost function parameters are varied. Debris avoidance maneuvers are considered, with the debris in the spacecraft rendezvous path.

Reference and command governors for systems with constraints

Reference and command governors are add-on control schemes which enforce state and control constraints on pre-stabilized systems by modifying, whenever necessary, the reference. This paper surveys the extensive literature concerning the development of such schemes for linear and nonlinear systems. The treatment of unmeasured disturbances and parametric uncertainties is also detailed. Generalizations, including extended command governors, feedforward reference governors, reduced order reference governors, parameter governors, networked reference governors, and decentralized/distributed reference governors, are discussed. Practical applications of these techniques are presented and surveyed as well. A comprehensive list of references is included. Connections with related approaches, including model predictive control and input shaping, are discussed. Opportunities and directions for future research are highlighted.