Abstract
This chapter provides a practical and intuitive way of cooperative task planning and execution for complex robotic systems using multiple robots in automated manufacturing applications. In large-scale complex robotic systems, because individual robots can autonomously execute their tasks, robotic activities are viewed as discrete event-driven asynchronous, concurrent processes. Further, since robotic activities are hierarchically defined, place/transition Petri nets can be properly used as specification tools on different levels of control abstraction. Net models representing inter-robot cooperation with synchronized interaction are presented to achieve distributed autonomous coordinated activities. An implementation of control software on hierarchical and distributed architecture is presented in an example multi-robot cell, where the higher level controller executes an activity-based global net model of task plan representing cooperative behaviors performed by the robots, and the parallel activities of the associated robots are synchronized without the coordinator through the transmission of requests and the reception of status.
TopIntroduction
Nowadays, advanced manufacturing systems are generally composed of the following control levels: planning, scheduling, coordination, and low-level control. At the planning level, based on definition of tasks and due dates, resource requirements are specified and an actual time period and resources are assigned to each task and subtask. The monitoring of real time events and modification and adaptation of the cell configuration after detection of an equipment failure are supported by a knowledge-based system. The coordination level is naturally decomposed into two sublevels: the flow of parts through the buffers and the coordination of a manufacturing operation. The former can be viewed as producer/consumer processes. The flow control and robot synchronization are performed by a multitasking operating system. The coordination level receives a rather static schedule from the upper level and is provided with dynamic information about particular states of a system from the lower level. It sends commands to low-level controllers and may ask for a rescheduling. Thus, the coordinated system behaves concurrently and includes cooperation and competition relations among its subsystems. This concurrency with links to both upper and lower levels should be modeled formally in the control system design. Currently, control systems for the coordination of specific low-level controllers according to the schedule of operations have been mainly developed in an ad hoc manner to fit just one particular system, which requires many man-years of hard programming again and again to build the control systems because specifications are not sufficiently complete and the system environment changes frequently.
The development of industrial techniques makes sequential operation control for robotic systems more large-scale and complicated one, in which some robots or subsystems operate concurrently and synchronously based on sensing information by external sensors and/or inter-robot communications as cooperative multi-robot systems (Yasuda et al., 1991). Multiple robots have the possibility to perform complex and large scale tasks more efficiently by cooperating in some way than a single robot does. Although there are many tasks that could be constructed for a single complex robot, in many cases there are advantages to using multiple robots, such as robustness and less complexity. However, to perform complex robotic tasks such as cooperative fixing and mating of separate parts by two or more robots, new distributed autonomous control architecture through the integration and cooperation for intelligent multi-robot tasks should be developed (Hörmann, 1991). In such intelligent robotic tasks, synchronization between associated robots is crucial, and cooperation based on ad hoc or temporary master-slave or leader-follower relationships or ideal cooperation through mutual coordination with synchronous communication based on simple reflexive behavioral control on a decentralized control architecture is required (Duffee, 1996). By specifying a conditional sequence of actions and goals or joint plan involving the subject and others, interlocking coordination where each robot adjusts its action selection based on the evolution of the ongoing interaction is achieved. In the distributed planning, the cooperative interaction consists of a series of actions including communication acts such that a robot causes a change in the environment that triggers the precondition of the next action of a sequence.
Key Terms in this Chapter
Hierarchical Planning: Planning formed and executed by several subsystems organized according to a layered structure for the system goal.
Centralized System: A system where decisions for the system goal are made in a central mechanism and transmitted to executive components.
Task Planning: Translation of a manufacturing plan into the robot operations that may allow the successful execution of the plan, involving the planning of sensory operations, gross-motion planning with collision avoidance, and fine-motion planning with contact motions.
Hierarchical Decomposition: Analysis of complex systems by replacing some parts of a model with simple components such that substitution nodes with their surrounding arcs are behaviorally equivalent to the related sub-models. In a discrete event net, either a place or a transition can be a substitution node by associating a sub-model to it.
Hierarchical Modeling: System modeling by describing a set of sub-models which all contribute to a much larger model by interacting with each other in a well-defined way.
Coordination: Decision making and their execution through negotiation processes among the executive components to fulfill a global goal in a distributed system.
Discrete Event Net Modeling: Modeling using a graphical net to capture the operational states and the events or transitions between them and plan a nominal operation sequence and recovery strategies due to uncertainties and tolerances.
Liveness: Possibility of any event to occur from any reachable state in a discrete event system. In a discrete event net, as deadlock arises when no transition can be longer enabled, liveness implies deadlock freedom.
Manufacturing Planning: Determination of a sequence of operations to be done in order to manufacture a product, such as machining or assembling.
Task Execution: Execution of robotic operations to carry out a manufacturing plan using code generation, involving concurrency, synchronization, resource sharing and real-time requirements.