PhD positions (3) on Integrated optimization-based and learning-based control of networks with hybrid dynamics

These 3 PhD projects are part of the European ERC Advanced Grant project CLariNet – a novel control paradigm for large-scale hybrid networks. The goal of CLariNet is to create a completely new paradigm for control of large-scale networks with hybrid dynamics by bridging the gap between optimization-based control and learning-based control. The breakthrough idea is to bridge that gap by using piecewise affine models and to unite the optimality of optimization-based control with the on-line tractability of learning-based control.

The 3 projects all have a strong fundamental flavor. In addition, applications for the case studies include multi-modal transportation networks and smart multi-energy networks.

Topic 1: Multi-agent integrated optimization-based and learning-based control for large-scale networks with hybrid dynamics

In this PhD project we will develop integrated optimization-based and learning-based control methods for large-scale hybrid systems – in particular piecewise affine (PWA) systems. More specifically, the aim is to develop several innovative approaches to combine model predictive control (MPC) and reinforcement learning so as to merge the advantages of both approaches, and to embed them in a distributed/multi-agent control setting. The main challenge will be to determine efficient approaches to obtain coordination among the control agents.

Topic 2: Integrated optimization-based and learning-based control for constrained hybrid systems in the presence of uncertainty

In this PhD project we will develop integrated optimization-based and learning-based control methods for hybrid systems – in particular piecewise affine (PWA) systems, in the presence of (stochastic) uncertainty and subject to input, output, and state constraints. The idea is to integrate scenario-based chance-constrained model predictive control (MPC) with learning-based control approaches. This also includes methods to efficiently obtain sets of representative scenarios that are rich enough so that performance guarantees can be given, and that be extracted in an efficient way from the huge amount of historical data that is available.

Topic 3: Performance analysis of integrated optimization-based and learning-based control for constrained hybrid systems

In this PhD project we will analyze and prove formal properties of integrated optimization-based and learning-based control methods for piecewise affine (PWA) systems subject to input, output, and state constraints. We will consider issues such as stability, computational complexity, error bounds, formal or probabilistic performance guarantees, robustness, finite termination effects, safety, etc. We will also investigate and characterize the various trade-offs (e.g., between allowed computation time and control performance/constraint violations).

The department Delft Center for Systems and Control (DCSC) of the faculty Mechanical, Maritime and Materials Engineering, coordinates the education and research activities in systems and control at Delft University of Technology. The Centers' research mission is to conduct fundamental research in systems dynamics and control, involving dynamic modelling, advanced control theory, optimisation and signal analysis. The research is motivated by advanced technology development in physical imaging systems, renewable energy, robotics and transportation systems.