---

Introduction

---

Outline of Components

The core of IDEA is a set of modules that define the foundation architecture and agent behaviors. IDEA is rooted strongly in interleaved planning and execution. The IDEA core does not presume a particular planning technology. That said, IDEA currently calls on a particular planning architecture called EUROPA. IDEA modules interface between the general core and EUROPA. Collectively, the core software and these adaptation modules form the IDEA-specific software base. In addition to this IDEA software base, the IDEA distribution includes EUROPA and a number of supporting, externally-developed software components.

IDEA

The IDEA core software defines an architecture that is common to all IDEA agents. In addition, it defines common API's for a number of functions that are common to all IDEA agents, but for which IDEA avoids making a commitment to a specific technology, for example the planning and scheduling method or communications protocols. In this case, IDEA also provides instantiations that a user can utilize for applications, and which are most widely-used approaches for IDEA developers. This includes instantiations for planning and scheduling (interfacing with EUROPA) and an instantiation of a communications protocol (utilizing IPC). IDEA is developed at NASA ARC.


For more information, please consult the below section on the IDEA Architecture.


EUROPA

EUROPA is a framework for enabling constraint-based planning and scheduling. It provides the means for defining a constraint database that propagates constraints as it acquires new knowledge, and determines whether the database continues to be consistent. On top of these tools and others, users can define planners that perform general or specific planning and scheduling tasks. As with IDEA, EUROPA provides instantiations of planners that can be used "off the shelf". IDEA uses two of these, the CBPlanner (Chronological Backtracking Planner, gradually being phased out) and the Solver (new planner that will be taking duties away from CBPlanner). EUROPA is also developed at NASA ARC.


External Components

IDEA also calls on a number of components generated outside NASA ARC, required when running IDEA. They include the following:
• IPC: A high-level communications protocol developed at Carnegie Mellon University by Reid Simmons and David Apfelbaum. It enables publish/subscribe and client/server type communications paradigms, and is used in a wide range of research applications due to its effectiveness and ease of use. IDEA uses IPC as a baseline communications protocol for inter-agent communications. For more information, please consult the CMU IPC site.
  
  
• Xerces: An XML parsing engine developed by the Apache XML Project. IDEA represents system models, initial state and heuristics data in terms of an XML-based language called XIDDL. Xerces allows models to be translated to other languages or even to be converted into data structures in high-level programming languages. For more information, please consult the Apache site.
  
  
• ACE-TAO: ACE (Adaptive Communication Network) is software that interfaces with a variety of operating systems but provides a consistent API for common operating system utilities. ACE makes it far easier to port IDEA between different platforms. TAO is a real-time CORBA implementation package that comes with the ACE distribution and integrates with ACE to provide a consistent high-level communications API. IDEA does not currently use TAO. For more information, please consult the ACE-TAO site.
  
  

IDEA Architecture

IDEA provides the same architecture for all agents. Its generic control structure provides several fundamental components that are defined in the IDEA core, and others that can be specialized for particular applications. The below figure illustrates the IDEA structure. IDEA core elements are shown in yellow and IDEA application-specific elements are shown in green. Both the yellow and green elements can be customized by the user. The gray elements are communications interfaces that are also part of the IDEA core.

IDEA Architecture Diagram

The IDEA components perform the following functions:


• PlanServer: Comprises the PslDatabase that stores the current planning state, and the PslModelManager, which imposes the constraints defined by domain models onto the propagation of information in the PslDatabase. Currently, the PslDatabase interfaces with a EUROPA temporal and parameter constraint database which maintains a constraint network given all incoming data, according to model rules. Importantly, all agent planners operate on the same database, guaranteeing consistency between planners.
  
  
• AgentTimingService: Acts as the interface between the agent and a globally-shared timing service reference. In addition, the AgentTimingService sets and enforces alarms set by the ReactivePlannerRunner in anticipation of scheduled events.
  
  
• ReactivePlanner: Plans over a single timing cycle to select and trigger events according to the current state of the planning database. The ReactivePlanner must restrict its planning cycle to a single latency time period, and typically plans over a time horizon of the same duration.
  
  
• ReactivePlannerRunner: Spawns messages for events and data passing as dictated by the ReactivePlanner, and sets wakeup alarms via the AgentTimingService in anticipation of scheduled events, as dictated by the planning database. The ReactivePlannerRunner allows the ReactivePlanner to sleep whenever there are neither scheduled events nor incoming messages over a "latency" period.
  
  
• AgentRelay: Handles message passing (for event transmission and data passing) within an agent and between agents. The AgentRelay uses CORBA as the underlying communications protocol.
  
  
• CommunicationsRelay: Translates between the IDEA-native communications of the AgentRelay and the agent-external communications or commanding protocols.
  
  
• GoalLoader: A specialized deliberative planner used to load pre-defined, externally-generated plans into the PslDatabase. Since plans are mostly or completely defined, the deliberative search is typically trivial.
  
  
• Reactors: Custom C++ modules activated by the ReactivePlanner to enact software associated with events that are represented in the PslDatabase. A common approach is to specialize a Reactor as a deliberative planner for long-term goal elaboration or plan generation within an agent.
  

Outline of Setup Process

IDEA setup is involved, but this tutorial is intended to guide users through the process step by step. The process breaks down into the following:
• Configuring Your System
• Getting the IDEA Software
• Building the Software
• Configuring IDEA for Running
• Running IDEA



The following section will get you started with configuring your system, getting the IDEA software, and building the software.

---

IDEA Tutorial Table of Contents Next: IDEA Setup

---

Contact information © IDEA