“…MAS can control several aspects of smart grids like management of energy, scheduling energy, reliability, the security of the network, fault handling capability, communication between agents (Mahela et al, 2020). Many real-time complex systems contain task execution dependencies (Lu, Nolte, Kraft, & Norstrom 2010), data dependencies (Ndoye & Sorel, 2013), and shared resources dependencies (Shi, Ueter, von der Brüggen, & Chen, 2019). Dependencies among tasks also need to be noted in scheduling tasks on a system of machines where the total energy consumed by the system is to be reduced.…”
We present a multi-agent system where agents can cooperate to solve a system of dependent tasks, with agents having the capability to explore a solution space, make inferences, as well as query for information under a limited budget. Re-exploration of the solution space takes place by an agent when an older solution expires and is thus able to adapt to dynamic changes in the environment. We investigate the effects of task dependencies, with highly-dependent graph G 40 (a well-known program graph that contains 40 highly interlinked nodes, each representing a task) and less-dependent graphs G 18 (a program graph that contains 18 tasks with fewer links), increasing the speed of the agents and the complexity of the problem space and the query budgets available to agents. Specifically, we evaluate trade-offs between the agent's speed and query budget. During the experiments, we observed that increasing the speed of a single agent improves the system performance to a certain point only, and increasing the number of faster agents may not improve the system performance due to task dependencies. Favoring faster agents during budget allocation enhances the system performance, in line with the "Matthew effect." We also observe that allocating more budget to a faster agent gives better performance for a less-dependent system, but increasing the number of faster agents gives a better performance for a highly-dependent system.
“…MAS can control several aspects of smart grids like management of energy, scheduling energy, reliability, the security of the network, fault handling capability, communication between agents (Mahela et al, 2020). Many real-time complex systems contain task execution dependencies (Lu, Nolte, Kraft, & Norstrom 2010), data dependencies (Ndoye & Sorel, 2013), and shared resources dependencies (Shi, Ueter, von der Brüggen, & Chen, 2019). Dependencies among tasks also need to be noted in scheduling tasks on a system of machines where the total energy consumed by the system is to be reduced.…”
We present a multi-agent system where agents can cooperate to solve a system of dependent tasks, with agents having the capability to explore a solution space, make inferences, as well as query for information under a limited budget. Re-exploration of the solution space takes place by an agent when an older solution expires and is thus able to adapt to dynamic changes in the environment. We investigate the effects of task dependencies, with highly-dependent graph G 40 (a well-known program graph that contains 40 highly interlinked nodes, each representing a task) and less-dependent graphs G 18 (a program graph that contains 18 tasks with fewer links), increasing the speed of the agents and the complexity of the problem space and the query budgets available to agents. Specifically, we evaluate trade-offs between the agent's speed and query budget. During the experiments, we observed that increasing the speed of a single agent improves the system performance to a certain point only, and increasing the number of faster agents may not improve the system performance due to task dependencies. Favoring faster agents during budget allocation enhances the system performance, in line with the "Matthew effect." We also observe that allocating more budget to a faster agent gives better performance for a less-dependent system, but increasing the number of faster agents gives a better performance for a highly-dependent system.
“…The applicability to complex systems with, e.g., error handling routines, may be questionable. In [Lu et al, 2010], the approach as presented in [Burns and Edgar, 2001] is further developed. The authors simulate the execution of a task in a typical embedded environment with state variables and scheduling.…”
Embedded real-time systems are growing in complexity, which goes far beyond simplistic closed-loop functionality. Current approaches of worst-case execution time (WCET) analysis are used to verify the deadlines of such systems, especially when they are safety-critical. These approaches calculate or measure WCET as a single value that is expected to be an upper bound for a system's execution time. Overestimations are taken into account to make this upper bound a safe bound, but modern processor architectures with caches, multi-threading, and instruction pipelines often expand those overestimations for safe upper bounds into unrealistic areas. Some approaches try to overcome this problem by calculating multiple upper bounds and argue that each single upper bound will hold for a certain probability (probabilistic worst-case execution time). Even though some of them tackle the problem of obtaining reliable probabilistic values for such upper bounds, more effort is required.Therefore, a method is presented in this thesis that combines probabilities of safety analysis models and elements of system development models in order to calculate a probabilistic worst-case execution time. Since safety analysis models are used to document the reliability or safety of safety-critical systems, they provide reliable probabilistic values. These probabilities are used here to calculate a probabilistic worst-case execution time that provides safe and reliable probabilities. The approach can be applied to systems that use mechanisms belonging to the area of fault tolerance, since such mechanisms are usually quantified in safety analyses to certify the system as being highly reliable or safe. A tool implementing this approach is also presented in this thesis. The tool provides reliable safe upper bounds by performing a static WCET analysis and overcomes the frequently encountered problem of dependence structures by using a fault injection approach. The tool can handle popular Simulink models and combines them with fault tree elements, a safety analysis model that is widely accepted by authorities, to derive probabilistic worst-case execution times. These execution times with probabilities provide information such as the probability of the system terminating within a given deadline.
“…To determine a probability distribution, there have been some initial approaches in probabilistic WCET analysis. In [10,11,12,13], the authors use measurement-based approaches. The central idea is to measure the timings of a task or response times and to stochastically drive probabilistic distributions for them.…”
The growing complexity of safety-critical embedded systems is leading to an increased complexity of safety analysis models. Often used fault tolerance mechanisms have complex failure behavior and produce overhead compared to systems without such mechanisms. The question arises whether the overhead for fault tolerance is acceptable for the increased safety of a system. Manually modeling the timing behavior is cost intensive and error prone. Current approaches of safety analysis and execution time analysis are not able to reflect the timing behavior of complex mechanisms according to failures. In this paper, we describe an approach that combines safety analysis models with execution times to extract different execution times for different failure conditions. This provides a detailed view on the safety behavior in combination with the produced overhead and allows to find and certify appropriate fault tolerance mechanisms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.