Integração on-line entre um sistema didático de manufatura controlado por CLP e um ambiente de planejamento automático hospedado em nuvem

Abstract

Programmable Logic Controllers (PLCs), although fundamental to traditional automation architectures, require considerable time for their programming development. The proposed approach aims to increase flexibility, modularity, and adaptability by using a high-level programming language. The work is based on PlanPAS, a developed solution that executes abstract actions (obtained from automated planners) directly in the field through the use of PLCs, sensors, actuators, and industrial communication networks. This approach is capable of integrating knowledge-based engineering (domain modeling, problem definition, and the generation of an action plan) with applied industrial automation (sensors, actuators, PLCs, and field data communication protocols). Despite the significant technical progress achieved at the time, PlanPAS operates in an offline manner, a non-intuitive methodology that limits operator adaptation to the system due to the need for manual reconfiguration of the plan. The proposed methodology is based on the implementation of an Application Programming Interface (API) developed in Python, capable of obtaining from planning.domains (an AI-planning platform) a solution file that specifies preconditions and describes all actions required to solve a given problem, which must be structured as representations of an initial scenario and a goal scenario. The planning problem is then formalized using information obtained directly from the field: the objectives to be achieved by the system are entered by the operator through an industrial interface, while the initial conditions of the process are constructed from the state of the sensors at the moment a new solution is requested. Within the proposed system, relevant information is exchanged at two distinct levels: from the API to the PLC, and vice versa, via an Industrial Ethernet network using the Modbus TCP protocol; and from the API to the AI-planning framework (PDDL Domains), and vice versa, via the HTTPS protocol using POST and GET methods. Physical process automation based on symbolic planning logic demonstrated: a flexible integration, tested on more than one PLC model supporting Modbus TCP communication; scalable, as demonstrated by domain expansion; and robust, presenting a contingency mechanism such that, in critical cases, the system remains operational even in the event of a temporary loss of communication with the API. This work enabled the integration between high-level decision-making processes, performed by an automated planner, and low-level execution, carried out by means of PLCs, which is characteristic of industrial applications.