Atef Gharbi

Model-checking for the functional safety of Control Component-based heterogeneous embedded systems

This paper1 deals with the model checking of Safe Heterogeneous Embedded Control Systems following different component-based technologies or implemented according to different Architecture Description Languages (ADL) used today in industry. The purpose is to reduce their time to market by exploiting various execution environments and different rich libraries. A “Control Component” is defined in our research work as an event-triggered software unit composed of an interface for any external interactions and an implementation allowing control actions of physical processes. A control system is assumed to be a composition of components with precedence constraints to control the plant according to well-defined execution orders. We define an agent-based architecture where the agent controls the environment evolution and applies automatic reconfigurations when hardware errors occur at run-time to guarantee a functional safety of the whole system. We model the architecture according to the formalism Net Condition/Event Systems (abbr. NCES), and apply the model checker SESA to check functional properties described according to the well-known Computation Tree Logic (abbr. CTL). Our purpose is to check that whenever an error occurs at run-time, the agent behaves as described in user requirements by activating Control Components and deactivating others to guarantee a functional safety of the whole system. A Benchmark Production System is used in this research work to explain our contribution.

Functional Safety of Component-based Embedded Control Systems

This paper deals with the development of Safe Heterogeneous Embedded Control Systems following different component-based technologies used today in industry to reduce their time to market. To cover these technologies, we define a “Control Component” as an event-triggered software unit supporting different control functionalities. To guarantee a functional safety of the system, we define an agent-based architecture where the agent controls the environment evolution and applies automatic reconfigurations when hardware errors occur at run-time. We model the whole architecture according to the formalism Net Condition/Event Systems (abbr. NCES), and apply the model checker SES A to verify properties described according to the Computation Tree Logic (abbr. CTL). Our purpose is to check that whenever an error occurs at run-time, the agent behaves as described in user requirements by activating control components and deactivating others in order to guarantee a functional safety of the whole system. A Benchmark Production System is used as a running example to explain our contribution.

The embedded control system through real-time task

This paper deals with the study of Embedded Control System through Real-Time Task. In the embedded control system, we decompose a component into several real-time tasks. To ensure safety control of tasks at run-time, we define service and reconfiguration processes for tasks and use the semaphore concept to obtain safety mutual exclusions. The Priority Ceiling Protocol (PCP) is applied as an approach to ensure the scheduling between periodic tasks with precedence and mutual exclusion constraints. We simulate the scheduling of Real-time tasks with PCP through the Cheddar tool.

Functional and Operational Solutions for Safety Reconfigurable Embedded Control Systems

The chapter deals with run-time automatic reconfigurations of distributed embedded control systems following component-based approaches. We classify reconfiguration scenarios into four forms: (1) additions-removals of components, (2) modifications of their compositions, (3) modifications of implementations, and finally (4) simple modifications of data. We define a new multi-agent architecture for reconfigurable systems where a Reconfiguration Agent which is modelled by nested state machines is affected to each device of the execution environment to apply local reconfigurations, and a Coordination Agent is proposed for any coordination between devices in order to guarantee safe and coherent distributed reconfigurations. We propose technical solutions to implement the whole agent-based architecture, by defining UML meta-models for agents. In the execution scheme, a task is assumed to be a set of components having some properties independently from any real-time operating system. To guarantee safety reconfigurations of tasks at run-time, we define service and reconfiguration processes for tasks and use the semaphore concept to ensure safety mutual exclusions. We apply the priority ceiling protocol as a method to ensure the scheduling between periodic tasks with precedence and mutual exclusion constraints.

Model checking optimization of safe Control Embedded Components with refinement

This paper deals with model checking optimization of Software Embedded Control Components by applying refinement. We introduce a Software Embedded Control Component as an event-triggered software unit composed of an interface for any external interactions and an implementation allowing control actions of physical processes. A control system is assumed to be a composition of components with precedence constraints to control the plant. To ensure safety, an intelligent software agent controls the plant and applies automatic reconfiguration whenever a physical error occurs in the plant. We propose in this paper to model checking these different reconfigurations through a refinement method realized step by step. The contributions of the paper are applied to two Benchmark Production Systems available in our research laboratory.

Optimal model checking of safe control embedded software components

The paper deals with reconfigurable software component-based embedded control systems. We define a Control Component as a software unit to support control tasks of the system which is assumed to be a network of components with precedence constraints. We define an agent-based architecture to handle automatic reconfigurations under well-defined conditions by creating, deleting or updating components to bring the whole system into safe and optimal behaviors. We model the agent by nested state machines where states correspond to other state machines in order to cover all reconfiguration forms. Several complex networks can implement the system where each one is executed at a given time when a corresponding reconfiguration scenario is automatically applied by the agent. To check the correctness of each one of them, we apply in several steps a refinement-based approach that automatically specifies feasible Control Components according to the formalism Net Condition/ Event Systems (abbr. NCES) which is an extension of Petri nets. The model checker SESA is automatically applied in each step to verify deadlock properties of new generated components, and it is manually used to verify CTL-based properties according to user requirements. Two Industrial Benchmark Production Systems FESTO and EnAS available in our research laboratory are applied to explain the paper contribution.

Reflective multi-agent model using semantic similarity measure and negotiation protocol for solving heterogeneity

Semantic similarity measures play important role in communication in an open and heterogeneous multi-agent system (MAS). A survey on similarity measures between concepts is afforded in this paper. We present these techniques, provide evaluations of their result performances, and discuss their shortcomings. We propose a measure by combining a psychological knowledge of the relevance, the resemblance and the non-symmetry of similarity. The proposed measure leads us to suggest a novel reflective agent model allowing agents to autonomously communicate between each other through semantic heterogeneity. The agent can enrich its own ontology by using semantic negotiation approach in several steps. We develop firstly, a model using an alignment ontology framework. Then, we improve a similarity measure to select the most similar pairs. Then, we suggest a protocol for supporting semantic negotiation. At the end, we present our experiments on many benchmark datasets proving that our results are more reasonable, we provide evaluations of some result performances of the existing semantic similarity metrics, and discuss their advantages and drawbacks.

Multi-agent Deep Reinforcement Learning for Task Allocation in Dynamic Environment.

The task allocation problem in a distributed environment is one of the most challenging problems in a multiagent system. We propose a new task allocation process using deep reinforcement learning that allows cooperating agents to act automatically and learn how to communicate with other neighboring agents to allocate tasks and share resources. Through learning capabilities, agents will be able to reason conveniently, generate an appropriate policy and make a good decision. Our experiments show that it is possible to allocate tasks using deep Q-learning and more importantly show the performance of our distributed task allocation approach.

Agent-based Semantic Negotiation Protocol for Semantic Heterogeneity Solving in Multi-agent System.

In this article, we propose an interactive agent model in an open and heterogeneous multi-agent system (MAS). Our model allows agents to autonomously communicate between each other through semantic heterogeneity. The communication problem can be expressed by the calculation based on the abilities acquired in the receiver agent, compared to the message sent by the sender agent. Hence, the semantic heterogeneity should be resolved in the message processing. The agent can autonomously enrich its own ontology by using semantic negotiation approach in several steps. We develop firstly, a model using an ontology alignment framework. Then, we enhance a similarity measure to select the most similar pairs by combining a psychological knowledge of the relevance, the resemblance, and the non-symmetry of similarity. At the end, we suggest a protocol for supporting semantic negotiation. In order to explain our approach, we implement a simple benchmark production system on JADE.

Five Capabilities Model Applied to Multi-Robot Systems

This paper deals with distributed Robotic Agents constituting several intelligent agents. Each one has to interact with the other autonomous robots. The problem faced is how to ensure a distributed planning through the cooperation of the distributed robotic agents. To do so, the author proposes to use the concept of five capabilities model which is based on Environment, Self, Planner, Competence, and Communication. A Benchmark Production System is used as a running example to explain the authors contribution.