MAST: Manufacturing Agent Simulation Tool
Brief description:
MASTMAST was developed in collaboration with Rockwell Automation as a demonstrator of key benefits of agent-based approaches to manufacturing control. Primary aim was at the simulation of dynamic product routing but over the years it matured into generic-purpose manufacturing simulation and control tool featuring real-time connectivity to legacy PLCs, ontology-based dynamic scheduling, advanced diagnostics, etc. It models the transportation system as a set of independent agents that apply dynamic path searching algorithm to discover optimal routing paths through the conveyor network. MAST is implemented in Java with the agent part programmed in the JADE open source agent platform.

DIAL photoThe first real application of MAST was established during collaboration with the DIAL laboratory at the University of Cambridge. MAST was modified to provide full-fledged agent-based simulation of the packing line producing customized gift boxes. The product agent was designed as an active entity that maintains state of an individual product and pursues the fulfillment of the order in terms of negotiations with resource agents about scheduling of packing operations. The behavior of agents is driven by the ontology, in which the description of the production process is described. This allows the product agent to automatically process the production plan of even new, previously unknown products (tests were performed with unpacking scenario). 

ACIN photoSecond deployment of MAST was aimed at simulating and also controlling the palette transfer system at the ACIN institute of Vienna University of Technology. MAST agents were provided with the access to real I/O values held in PLC in order to sense and actuate in the real environment. Under real conditions we have verified the ability of agents to dynamically reconfigure the control system in terms of finding and applying alternative routing in case of a conveyor failure.

CWS: Chilled Water System
CWSCWS is a multi-agent solution developed in cooperation with Rockwell Automation for real-time control of US Navy ship’s cooling system. The key requirement was provide a solution that ensures cooling system functionality in cases of severe failures or damages of the equipment. The agent-based system was designed, in which each onboard equipment, such as cooling unit, valve and load, is governed by an autonomous agent. Agents dynamically negotiate about provisioning of cold water and its transport through the redundant piping system (see Fig. 1). A unique feature is the ability to detect and isolate leakage and subsequently to reconfigure the control system in terms of finding the alternative routing of water in order to continue in cooling of the critical ship systems. The control system was deployed and sucessfully tested on the US Navy's Reduced Scale Advanced Development model serving as reconfigurable fluid system test platform.

Passive House Simulation & Control
Passive houses Passive House Simulation & Control are featured with a very low energy consumption for heating and operation. The low energy consumption cannot be met with decreased indoor air quality or living comfort, hence there should be used proper control algorithms to find the optimum in such a multi-variable and non-linear control problem.

The goal of this project is to design and verify a modular simulation model that can be used as a test-bed for analyzing the quality of control algorithms and for fine-tuning their constants. The simulation model is currently being compared with measured data from a real passive house solving as a test-case. A set of testing scenarios, which has been performed in the experimental house, is intended not only to prove the particular model, but also to obtain results, which are valid for the whole class of passive family houses.

In the experimental passive house, the deployment of the project outcomes includes the wireless measuring system, a PLC-based solution for sunblind and lighting control, and a smart HMI that visualizes runtime data on an embedded touch screen, logs time-series of all measured data, and enters user commands into the control system.