Andrés Terrasa

Analyzing the schedulability of hard real-time artificial intelligence systems☆

Abstract This paper presents a prototype of the architecture known as the Architecture for Real-Time Intelligent Systems (ARTIS). Since the aim of this work is to have a tool for developing hard real-time intelligent systems, the theoretical fixed pre-emptive schedulability equations have been extended by including the prototypes exact timing factors. Some architectural design decisions, as well as the behaviour of an event-driven, non-preemptable, real-time kernel, are also presented, in order to include their effects in these schedulability equations.

TRAMMAS: A tracing model for multiagent systems

Agents flexibility and autonomy, as well as their capacity to coordinate and cooperate, are some of the features which make multiagent systems useful to work in dynamic and distributed environments. These key features are directly related to the way in which agents communicate and perceive each other, as well as their environment and surrounding conditions. Traditionally, this has been accomplished by means of message exchange or by using blackboard systems. These traditional methods have the advantages of being easy to implement and well supported by multiagent platforms; however, their main disadvantage is that the amount of social knowledge in the system directly depends on every agent actively informing of what it is doing, thinking, perceiving, etc. There are domains, for example those where social knowledge depends on highly distributed pieces of data provided by many different agents, in which such traditional methods can produce a great deal of overhead, hence reducing the scalability, efficiency and flexibility of the multiagent system. This work proposes the use of event tracing in multiagent systems, as an indirect interaction and coordination mechanism to improve the amount and quality of the information that agents can perceive from both their physical and social environment, in order to fulfill their goals more efficiently. In order to do so, this work presents an abstract model of a tracing system and an architectural design of such model, which can be incorporated to a typical multiagent platform.

Extracting temporal properties from real-time systems by automatic tracing analysis

Statically analyzing real-time systems normally involves a high degree of pessimism, but it is necessary in systems requiring 100% guarantee. However, lots of less critical systems would significantly benefit from combining such static analysis with empirical tests. Empirical tests are based on observing the system at run time and extracting information about its temporal behavior. In this sense, this paper presents a generic and extensible framework that permits the extraction of temporal properties of real-time systems by analyzing their run-time traces. The analysis is based on event-recognition finite state machines that compute the temporal properties with a computational cost of O(1) per observed event in most of the cases. The framework is instantiated in order to extract some typical temporal properties (such as computation time or response time of tasks), which can serve as a template to define new ones. Finally, the paper also shows how the framework can be implemented on a real system, exclusively using state-of-the-art technology; in particular, the Trace and Real-Time Extensions of the POSIX standard.

A meta-reasoning model for hard real-time agents

This paper defines a meta-reasoning model for real-time agents. The purpose of this model is to increase the adaptability in this kind of agents. This model allows to adapt the agents behavior as well as the agents own reasoning process. The application of this model to an specific real-time agent architecture is also presented.

Real-time synchronization between hard and soft tasks in RT-Linux

This paper presents the experience in adding a synchronization protocol to our architecture for flexible hard real-time systems. The architecture combines mandatory and optional components in a two-level tasking model. Mandatory and optional components share resources, and therefore synchronization is needed in order to maintain both the logical correctness of the resources and the timing correctness of the hard task set. Different approaches were analyzed and finally we implemented and evaluated both the Priority Ceiling Protocol (PCP) and the Ceiling Semaphore Protocol (CSP) within our run-time support system. Our results show that both methods can be made to work, but that the CSP method is generally more efficient and scales better.

A Tracing System Architecture for Self-adaptive Multiagent Systems

This paper proposes an event trace architecture, based in the concept of service, which can be used to increase the amount and quality of the information that agents perceive from both their physical and social environments in order to know when to trigger a reorganisation.

Engineering a Tool for Building Hard Predictable Real-Time Artificial Intelligent Systems 1

Abstract In this paper we present a prototype of our architecture called an Architecture for Real-Time Intelligent Systems (ARTIS). Since our aim is to have a tool for developping hard real-time intelligent systems, we have extended the theoretical fixed pre-emptive schedulability equations by including the prototypes exact timing factors. Some architectural design decisions as well as the behaviour of an event-driven, non-preemptable, real-time kernel are also presented, in order to include their effects in these schedulability equations.

Flexible Real-Time Linux*: A Flexible Hard Real-Time Environment

This paper presents a framework appropriate for Flexible Real-Time Systems (FRTS) and a run-time support system based on that framework, called Flexible Real-Time Linux (FRTL). The framework proposes to build each task as a sequence of mandatory and optional components and to separate their execution in two scheduling levels. This approach is shown to provide both hard guarantees and flexible behavior. The FRTL system has been implemented by enhancing the original capabilities of Real-Time Linux (RT-Linux), while maintaining its predictability and efficiency features. This paper also shows a complete schedulability test on which all sources of overhead of the FRTL itself have been introduced. By applying this complete test, the designer is able to safely guarantee a real-time application running on the FRTL system.

Extending RT-Linux to Support Flexible Hard Real-Time Systems with Optional Components

This paper describes a framework that provides a task programming model with optional components and the appropriate operating system mechanisms for supporting it. The approach provides 100% guarantees to hard real-time tasks using fixed priority pre-emptive scheduling. Optional components, which increase the utility of the guaranteed tasks, are executed when spare processor capacity is available. The framework has been developed extending the existing RT-Linux capabilities. Furthermore, the design of these kernel extensions have been proved to be predictable, in such a way that it is possible to perform a schedulability analysis of the entire system, including kernel overheads.

A Predictable Module for Sharing Memory in RT-LINUX 1

Abstract Recently, RT-Linux has emerged as a promising approach for developing hard real-time systems. RT-Linux is a real-time executive placed underneath the Linux kernel. This executive basically provides a support for running real-time tasks at the kernel level and turns the whole Linux into one of these tasks. This paper relates to extending RT-Linux in order to have a way to share memory between real-time tasks and Linux user processes belonging to a real-time application. The shared memory extension (implemented in two separate modules) basically provides a storage manager, which allows dynamic allocation (and deallocation) of memory objects. The interface functions provided by this manager are shown to be predictable.