François Guerrin

Qualitative reasoning about an ecological process: interpretation in hydroecology

Abstract Because of the frequent inefficiency of classical mathematical modelling to help the human operators in the supervision of biological processes, we present here a method based on qualitative reasoning concepts for simulating the interpretation of measurements, analyses, and observations, commonly done on aquatic ecosystems for management purposes. Once the domain variables are identified, their cause-effect dependences are represented as a directed graph. Each variable takes its value in a five-symbol set called quantity space (QS). These symbols … pp, p, m, f, ff…, correspond to expert qualifiers, like respectively: very low, low, medium, high, very high. A synthetic formalism is proposed to encode four types of knowledge rules: (1) translation of the numerical input values (i.e. measurements and results of analyses) into qualitative values; (2) translation of the linguistic observations into qualitative values; (3) formal calculus on QS, using six empirically defined operators, for propagating a top-down form of reasoning throughout the causal network, enabling the determination of the qualitative values of unmeasured variables from the values of their causes; (4) control of the execution of the reasoning. The software prototype, implemented in Prolog, has four main functions: short-term prediction of management parameters, causal explanation of the reasoning, state memorization, and choice of control variables in the causal network. These capabilities are illustrated by examples from an application enabling the interpretation of data in hydroecology. The relevance of a qualitative reasoning approach is emphasized, particularly for making empirical knowledge, typical of biological process control, operational.

Knowledge representation and qualitative simulation of salmon redd functioning. Part I : Qualitative modeling and simulation

This work aims at representing empirical knowledge of freshwater ecologists on the functioning of salmon redds (spawning areas of salmon) and its impact on mortality of early stages. For this, we use Qsim, a qualitative simulator. In this first part, we provide unfamiliar readers with the underlying qualitative differential equation (QDE) ontology of Qsim: representing quantities, qualitative variables, qualitative constraints, QDE structure. Based on a very simple example taken of the salmon redd application, we show how informal biological knowledge may be represented and simulated using an approach that was first intended to analyze qualitatively ordinary differential equations systems. A companion paper (Part II) gives the full description and simulation of the salmon redd qualitative model. This work was part of a project aimed at assessing the impact of the environment on salmon populations dynamics by the use of models of processes acting at different levels: catchment, river, and redds. Only the latter level is dealt with in this paper.

Knowledge representation and qualitative simulation of salmon redd functioning. Part II : Qualitative model of redds

This paper describes a qualitative model of the functioning of salmon redds (spawning areas of salmon) and its impact on mortality rates of early stages. For this, we use Qsim, a qualitative simulator, which appeared adequate for representing available qualitative knowledge of freshwater ecology experts (see Part I of this paper). Since the number of relevant variables was relatively large, it appeared necessary to decompose the model into two parts, corresponding to processes occurring at separate time-scales. A qualitative clock allows us to submit the simulation of salmon developmental stages to the calculation of accumulated daily temperatures (degree-days), according to the clock ticks and a water temperature regime set by the user. Therefore, this introduces some way of real-time dating and duration in a purely qualitative model. Simulating both sub-models, either separately or by means of alternate transitions, allows us to generate the evolutions of variables of interest, such as the mortality rates according to two factors (flow of oxygenated water and plugging of gravel interstices near the bed surface), under various scenarios.

MAGMA: a simulation model to help manage animal wastes at the farm level

MAGMA is a hybrid dynamical system (HDS; i.e. with both continuous and discrete-valued variables) allowing the simulation of management of various kinds of animal manure or slurry production and utilization modes (waste spreading on cultivated crops and fallow land and compost making). It can support decision-making to help better manage such potentially hazardous organic matter (OM) at the farm level with the aims of environmental risk minimization, agricultural efficiency and farming sustainability. One can define several scenarios based on farm structural parameters in MAGMA, for example, animal and crop production characteristics, capacity of manure or slurry spreading equipment, and distances. Each scenario can be tailored according to several management strategies by using switch constants. This paper describes the core of the model, enabling the simulation and scheduling of actions (crop harvest, manure or slurry spreading, etc.) that consume those wastes according to various uses. The use of the model in simulation is explained and a detailed example is given on how it can contribute in choosing alternative management strategies.

Dynamic simulation of action at operations level

The attempt of using lumped or agent-based simulation models to support operations management in production systems puts action modelling to the fore. To fill the gap of classical decision-support systems ignoring human agents’ practices, a modelling framework of action at operations level is proposed. This framework aims at answering two questions: How to represent action? How to represent the management of action? Every action (i.e., what is actually done by an agent) is represented by a binary function of time governed by events detected upon processes of various kinds: artefacts (clocks or schedules), external processes occurring in the environment, other actions. In turn, every action exerts its effect on target processes. This modelling framework allows one to simulate the interpretation of ongoing actions by using temporal or propositional logics and operations management behaviors through plan specification and execution, action composition, and resource allocation to concurrent actions. It enables complex activity systems to be represented and management options to be tested by simulation. These capacities are illustrated on the example of a farming system. The main benefits and issues raised by this dynamical system approach close to the ‘situated’ (vs. ‘planned’) action paradigm are discussed in the light of related works in Artificial intelligence. Future directions of research are drawn, namely that of how to scale up this lower-level representation of action to the higher-level representation of agents endowed with skills relevant at the level of the individual (e.g., anticipation).

Managing the Supply of a Slurry Treatment Plant by Means of a Hybrid Dynamical System

Abstract A way to deal with the problem of surpluses of animal wastes is to implement collective treatment units. With aim to avoid stock overflowing and get some guarantee on the regularity of supply, this paper considers two questions: When should a delivery be done? How much has to be delivered? Based on an Hybrid Dynamical System approach (i.e. with continous and discrete-valued variables), we developed a model to answer these questions by simulating various management policies and optimizing control parameters. After describing this model, we show on an hypothetical case, involving 5 pig farms and 3 management policies, how it can be used to determine, for each farm, the dates and quotas of slurry deliveries and, further, to improve the overall system organization

Abstraction of continuous system trajectories into timed automata

Abstract This paper deals with the representation of continuous system dynamics into a timed discrete-event formalism to the end of system analysis. The continuous model of the system is first approximated using intervals and then translated into the timed automata formalism by comparison with thresholds defined on the state variables’ domains. The detection of thresholds crossing is characterised by two time instants corresponding respectively to the earliest and latest crossing dates. This approach is briefly illustrated with real data obtained from a 1 m wastewater treatment pilot plant.

MAGMA: A Model to Help Animal Manure Management at the Farm Level

Abstract MAGMA is an hybrid dynamical system allowing one to simulate various kinds of animal manure or slurry production and consumption modes (spreading on cultivated crops and fallow land, compost making). It can support decision-making to help better manage such potentially hazardous organic matter at the farm level with aims environmental risk minimisation, agricultural efficiency and farming sustainability. This paper emphasises the use of the model in simulation and describes a detailed example of how it can contribute to choosing among alternative management strategies.

Simulation of Actions to Help Animal Wastes Management at the Farm Level

Abstract MAGMA is an hybrid dynamical system allowing one to assess, at the farm level, the impact of various management scenarios on animal manure or slurry stocks in terms of agricultural benefits, environmental risks and labor, Here is described the core of the model, enabling the simulation and scheduling of actions (crop harvest, manure or slurry spreading) that consume those wastes according to various uses.

Qualitative reasoning methods for CELSS modeling

Qualitative Reasoning (QR) is a branch of Artificial Intelligence that arose from research on engineering problem solving. This paper describes the major QR methods and techniques, which, we believe, are capable of addressing some of the problems that are emphasized in the literature and posed by CELSS modeling, simulation, and control at the supervisory level.