Motion control systems with unconventional actuators based on smart materials
This activity concerns the research on motion control and vibration damping using devices base on electrically or magnetically active materials (also referred to as smart materials). The research on this subject is made in collaboration with the teams of professors Hartmut Janocha and Stefan Seelecke of the Dept. of Mechatronics, University of Saarland, Saarbruecken, Germany. The research involves the design of innovative control laws to deal with the strong non-linearities and uncertainties introduced by the electrically or magnetically active element (such as hysteresis, stick-slip or other internal friction effects) or by the complex mechanical strutture that is necessary to perform mechanical amplification of motion or force. The research activity is focused on the combination of adaptive and robust control tools in conjunction with advanced hysteresis or friction modeling tools.
Nonlinear control of electrical drives for precise positioning systems
This research regards the development of advanced design and optimization methods for robotics, industrial drives, and other embedded control systems. This research line investigates new adaptive methods based on the incorporation of computationally-light stochastic optimization algorithms (such as the family of simultaneous perturbation stochastic gradient approximation methods, or the compact genetic algorithms) in advanced control schemes (e.g., vectorial control of induction motors with multiple feedback loops, Lyapunov-based direct and indirect adaptive control schemes) with particular emphasis on the actual implementation in commercially-available microcontrollers for embedded systems.
Modeling, simulation, and distributed control of large-scale discrete-event systems
The recent advances in information and communication technologies have significantly influenced the evolution of control architectures for large-scale systems (manufacturing systems, warehouses, wireless/mobile sensor networks), generating a shift from traditional centralized and hierarchical schemes to distributed networks of low-cost, versatile sensor/actuator units governed by autonomous controllers (agents). The design of appropriate decision and control algorithms for each autonomous node, and of coordination strategies capable to lead the whole network to the desired global behavior involves many challenging research issues. In this context, the research efforts are focused on the problem of agent modeling and on the optimization of coordination strategies. The agent modeling is based on formal discrete-event modeling tools, such as the Discrete EVent System specification (DEVS) approach developed by B.P. Ziegler, or the matrix-based formulation for discrete event control developed by Frank Lewis. The coordination of agents is mainly addressed with bio-inspired optimization algorithms from the class of Evolutionary Computation methods.
Computational Intelligence for control and fault diagnosis of industrial processes
This research activity deals with the development and prototyping of intelligent sensors for monitoring, fault diagnosis and control of industrial processes. In this context, the recent research studies are mainly focused on innovative algorithms based on Computational Intelligence for laser sensor signal processing. Recent successful applications include an advanced sensor for weld-defect detection, which measures the radiations emitted by the plasma surrounding the welding arc, and processes the information in real time (through Kalman filtering and fuzzy data fusion) to determine an index of local quality of the weld. Another effective outcome of this research line is the development of a fuzzy filter for the detection and removal of spike noise in laser-based automated railway monitoring system.
Scheduling and dispatching of production activities
The automation of production activities in manufacturing systems involves a variety of planning, decision and optimization problems over different time horizons. This research line consists of a set of partially interrelated studies dedicated to the development of effective meta-heuristic optimization algorithms for activity scheduling, rescheduling and dispatching. With regards to scheduling problems, a number of effective search algorithms has been developed combining bio-inspired stochastic search tools (such as Evolutionary Algorithms) with effective constructive heuristics capable to quickly refine the known solutions. This approach has been applied to industrial case studies of challenging complexity, which include batch production of furniture, and the production and distribution of rapidly perishable goods. Similarly, in the context of dispatching problems (e.g., the control of automated guided vehicles for material handling), an approach based on the combination of multi-criteria decision rules with combinatorial “look-ahead” algorithms has been developed to overcome the inherent myopia of the conventional strategies generally adopted in this context.