Agent logiciel | EPFL Graph Search

En informatique, un agent ou agent logiciel (du latin agere : agir) est un logiciel qui agit de façon autonome. C'est un programme qui accomplit des tâches à la manière d'un automate et en fonction de ce que lui a demandé son auteur. Dans le contexte d'Internet, les agents intelligents sont liés au Web sémantique, dans lequel ils sont utilisés pour faire à la place des humains les recherches et les corrélations entre les résultats de ces recherches. Ceci se fait en fonction de règles prédéfinies. Ils sont capables d'une certaine autonomie, en particulier de dialoguer entre eux. Par exemple, l'agent intelligent d'une personne qui souhaite faire un achat sera capable de dialoguer avec les agents des vendeurs pour comparer prix, qualité et prestations. Par ailleurs, l'étude des interactions entre agents, lorsqu'ils sont plusieurs, est le domaine spécifique des systèmes multi-agents, discipline de l'informatique apparue dans les années 1980. Les agents mobiles La programmat

Reliability Mechanisms for Controllers in Real-Time Cyber-Physical Systems

Maaz Mashood Mohiuddin

Cyber-physical systems (CPSs) are real-world processes that are controlled by computer algorithms. We consider CPSs where a centralized, software-based controller maintains the process in a desired state by exchanging measurements and setpoints with process agents (PAs). As CPSs control processes with low-inertia, e.g., electric grids and autonomous cars, the controller needs to satisfy stringent real-time constraints. However, the controllers are susceptible to delay and crash faults, and the communication network might drop, delay or reorder messages. This degrades the quality of control of the physical process, failure of which can result in damage to life or property. Existing reliability solutions are either not well-suited for real-time CPSs or impose serious restrictions on the controllers. In this thesis, we design, implement and evaluate reliability mechanisms for real-time CPS controllers that require minimal modifications to the controller itself. We begin by abstracting the execution of a CPS using events in the CPS, and the two inherent relations among those events, namely network and computation relations. We use these relations to introduce the intentionality relation that uses these events to capture the state of the physical process. Based on the intentionality relation, we define three correctness properties namely, state safety, optimal selection and consistency, that together provide linearizability (one-copy equivalence) for CPS controllers. We propose intentionality clocks and Quarts, and prove that they provide linearizability. To provide consistency, Quarts ensures agreement among controller replicas, which is typically achieved using consensus. Consensus can add an unbounded-latency overhead. Quarts leverages the properties specific to CPSs to perform agreement using pre-computed priorities among sets of received measurements, resulting in a bounded-latency overhead with high availability. Using simulation, we show that availability of Quarts, with two replicas, is more than an order of magnitude higher than consensus. We also propose Axo, a fault-tolerance protocol that uses active replication to detect and recover faulty replicas, and provide timeliness that requires delayed setpoints be masked from the PAs. We study the effect of delay faults and the impact of fault-tolerance with Axo, by deploying Axo in two real-world CPSs. Then, we realize that the proposed reliability mechanisms also apply to unconventional CPSs such as software defined networking (SDN), where the controlled process is the routing fabric of the network. We show that, in SDN, violating consistency can cause implementation of incorrect routing policies. Thus, we use Quarts and intentionality clocks, to design and implement QCL, a coordination layer for SDN controllers that guarantees control-plane consistency. QCL also drastically reduces the response time of SDN controllers when compared to consensus-based techniques. In the last part of the thesis, we address the problem of reliable communication between the software agents, in a wide-area network that can drop, delay or reorder messages. For this, we propose iPRP, an IP-friendly parallel redundancy protocol for 0 ms repair of packet losses. iPRP requires fail-independent paths for high-reliability. So, we study the fail-independence of Wi-Fi links using real-life measurements, as a first step towards using Wi-Fi for real-time communication in CPSs.

EPFL2018

Reliable and Robust Cyber-Physical Systems for Real-Time Control of Electric Grids

Wajeb Saab

Real-time control of electric grids is a novel approach to handling the increasing penetration of distributed and volatile energy generation brought about by renewables. Such control occurs in cyber-physical systems (CPSs), in which software agents maintain safe and optimal grid operation by exchanging messages over a communication network. We focus on CPSs with a centralized controller that receives measurements from the various resources in the grid, performs real-time computations, and issues setpoints. Long-term deployment of such CPSs makes them susceptible to software agent faults, such as crashes and delays of controllers and unresponsiveness of resources, and to communication network faults, such as packet losses, delays, and reordering. CPS controllers must provide correct control in the presence of external non-idealities, i.e., be robust, and in the presence of controller faults, i.e., be reliable. In this thesis, we design, test, and deploy solutions that achieve these goals for real-time CPSs. We begin by abstracting a CPS for electric grids into four layers: the control layer, the network layer, the sensing and actuation layer, and the physical layer. Then, we provide a model for the components in each layer, and for the interactions among them. This enables us to formally define the properties required for reliable and robust CPSs. We propose two mechanisms, Robuster and intentionality clocks, for making a single controller robust to unresponsive resources and non-ideal network conditions. These mechanisms enable the controller to compute and issue setpoints even when some measurements are missing, rather than to have to wait for measurements from all resources. We show that our proposed mechanisms guarantee grid safety and outperform state-of-the-art alternatives. Then, we propose Axo: a framework for crash- and delay-fault tolerance via active replication of the controller. Axo ensures that faults in the controller replicas are masked from the resources, and it provides a mechanism for detecting and recovering faulty replicas. We prove the reliable validity and availability guarantees of Axo and derive the bounds on its detection and recovery time. We showcase the benefits of Axo via a stability analysis of an inverted pendulum system. Solutions based on active replication must guarantee that the replicas issue consistent setpoints. Traditional consensus-based schemes for achieving this are not suitable for real-time CPSs, as they incur high latency and low availability. We propose Quarts, an agreement mechanism that guarantees consistency and a low bounded latency- overhead. We show, via extensive simulations, that Quarts provides an availability at least an order of magnitude higher than state-of-the-art solutions. In order to test the effect of our proposed solutions on electric grids, we developed T-RECS, a virtual commissioning tool for software-based control of electric grids. T-RECS enables us to test the proper functioning of the software agents both in ideal and faulty conditions. This provides insight into the effect of faults on the grid and helps us to evaluate the impact of our reliability solutions. We show how our proposed solutions fit together, and that they can be used to design a reliable and robust CPS for real-time control of electric grids. To this end, we study a CPS with COMMELEC, a real-time control framework for electric grids via explicit power setpoints. We analyze the reliability issues...

EPFL2019

T-RECS: A Software Testbed for Multi-Agent Real-Time Control of Electric Grids

Jagdish Prasad Achara, Jean-Yves Le Boudec, Roman Rudnik, Wajeb Saab

Multiple software agents can be used to perform the real-time control of electrical grids. The control performance of such solutions is influenced by software non-idealities such as crashes and delays of the software agents, and message losses and delays due to the underlying communication network. To study the effect of these non-idealities on control systems, we present an open-source software testbed, named T-RECS. It uses software containers to test existing software without modification. The communication network among the software containers is emulated using Mininet framework, which allows for real packets being exchanged. The electric resources in the grid are simulated using state-of-the-art models, whereas the grid itself is modeled in the phasor domain. As control agents are run as is and message exchanges are emulated, T-RECS accurately captures the real-world properties of the control framework. We demonstrate the working of T-RECS with the Commelec control framework and show the effect of network non-idealities on the control performance. We make a beta version available.

2017