In this paper, the reasoning and planning capabilities of an autonomous mobile robot are described. The reasoning system that controls the robot is designed to exhibit the kind of behavior expected of a rational agent, and is endowed with the psychological attitudes of belief, desire, and intention. Because these attitudes are explicitly represented, they can be manipulated and reasoned about, resulting in complex goal-directed and reflective behaviors. Unlike most planning systems, the plans or intentions formed by the robot need only be partly elaborated before it decides to act. This allows the robot to avoid overly strong expectations about the environment, overly constrained plans of action, and other forms of overcommitment common to previous planners. In addition, the robot is continuously reactive and has the ability to change its goals and intentions as situations warrant. The system has been tested with SRI's autonomous robot (Flakey) in a space station scenario involving navigation and the performance of emergency tasks.

Reactive reasoning and planning

Most artificial intelligence research investigates intelligent behavior for a single agent--solving problems heuristically, understanding natural language, and so on. Distributed Artificial Intelligence (DAI) is concerned with coordinated intelligent behavior: intelligent agents coordinating their knowledge, skills, and plans to act or solve problems, working toward a single goal, or toward separate, individual goals that interact. DAI provides intellectual insights about organization, interaction, and problem solving among intelligent agents. This comprehensive collection of articles shows the breadth and depth of DAI research. The selected information is relevant to emerging DAI technologies as well as to practical problems in artificial intelligence, distributed computing systems, and human-computer interaction. 

"Readings in Distributed Artificial Intelligence" proposes a framework for understanding the problems and possibilities of DAI. It divides the study into three realms: the natural systems approach (emulating strategies and representations people use to coordinate their activities), the engineering/science perspective (building automated, coordinated problem solvers for specific applications), and a third, hybrid approach that is useful in analyzing and developing mixed collections of machines and human agents working together. 

The editors introduce the volume with an important survey of the motivations, research, and results of work in DAI. This historical and conceptual overview combines with chapter introductions to guide the reader through this fascinating field. A unique and extensive bibliography is also provided.

Readings in Distributed Artificial Intelligence

To date, reactive robot behavior has been achieved only through manual programming. This paper describes a new kind of plan, called a "universal plan", which can be synthesized automatically, yet generates appropriate behavior in unpredictable environments. In classical planning work, problems were posed with unique initial and final world states; in my approach a problem specifies only a goal condition. The planner is thus unable to commit to any specific future course of events but must specify appropriate reactions for anticipated situations. An alternative conception is that one universal plan compactly represents every classical plan. Which part of the universal plan is executed depends entirely on how the environment behaves at execution time.

Universal plans are constructed from state-space operator schemas by a nonlinear planner. They explicitly identify predicates requiring monitoring at each moment of execution, and provide for sabotage, serendipity and failure without requiring replanning.

/pdf/universal-plans-for-reactive-robots-in-unpredictable-4c93m3kzdo.pdf

Universal plans for reactive robots in unpredictable environments

The Procedural Reasoning System, a generic reasoning system based on the rational-agent architecture, is described. The system has been applied to several real-time applications, including mobile robot control, system control for a surveillance aircraft, and air traffic management. Two applications that are relevant to process industries handling malfunctions for the reaction control system of NASA's Space Shuttle and diagnosing and controlling failures and overloads in a telecommunications network are discussed. Specifically, the representation and management of malfunction-handling procedures and tasks. >

https://hal.laas.fr/hal-01980099/document

An architecture for real-time reasoning and system control

A robot acting in the real world must use flexible plans because actions will sometimes fail to produce desired effects, and unexpected events will sometimes demand the robot shift its attention. A plan is usually construed as a list of primitive robot actions to be executed one after another but in a complex domain, a plan must be structured to cope effectively with the myriad unpredictable details it will encounter during execution. However, adding structure to a plan involves more than augmenting the primitive plan representation; it requires a complete model of interaction with the world called situation-driven execution. Situation-driven execution assumes that a plan consists of tasks with three major components: a satisfaction test, a window of activity, and a set of execution methods that are appropriate in different circumstances. Execution of such a plan proceeds by selecting an unsatisfied task and choosing a method to achieve it based on the current world state. A task may be executed as many times as necessary to keep it satisfied while it is active.
This thesis proposes a plan and task representation based on program-like reactive action packages, or RAPs. A plan consists of RAP-defined tasks and each RAP generates primitive robot actions at execution time by selecting its most appropriate method. Within such a system, execution monitoring becomes an intrinsic part of the execution algorithm, and the need for separate replanning on failure disappears. RAPs are more than just programs that run at execution time, however, they are also hierarchical building blocks for plan construction. The RAP representation is structured to make a task's expected behavior evident for use in planning as well as in execution. The RAP execution system described includes a sensor memory, representation language and interpreter. Examples and experiments demonstrate a wide range of adaptive system behavior.

Adaptive execution in complex dynamic worlds

: In this paper, the reasoning and planning capabilities of an autonomous mobile robot are described. The reasoning system that controls the robot is designed to exhibit the kind of behavior expected of a rational agent, and is endowed with the psychological attitudes of belief, desire, and intention. Because these attitudes are explicitly represented, they can be manipulated and reasoned about, resulting in complex goal-directed and reflective behaviors. Unlike most planning systems, the plans or intentions formed by the robot need only be partly elaborated before it decides to act. This allows the robot to avoid overly strong expectations about the environment, overly constrained plans of action, and other forms of overcommitment common to previous planners. In addition, the robot is continuously reactive and has the ability to change its goals and intentions as situations warrant. Thus, while the system architecture allows for reasoning about the means and ends in much the same way as traditional planners, it also possesses the reactivity required for survival in highly dynamic and uncertain worlds. The system has been tested with SRI's autonomous robot (Flakey) in a space station scenario involving navigation and the performance of emergency tasks.

/pdf/reasoning-and-planning-in-dynamic-domains-an-experiment-with-1j0v3vzqvh.pdf

Marcel J. Schoppers

Papers

Reasoning and planning in dynamic domains: An experiment with a mobile robot