Combining Classical and Reactive Planning: the ABC Model

Abstract

Systems that must plan, act, and probably learn using a rational, flexible behavior in real time present a challenge to current integrated schemas of planning and acting. Classical planning is still far from solving problems on such short time as needed in dynamic domains, such as Robosoccer, even with new approaches such as graph-based techniques. Reactive systems on the other hand, are hard to configure and do not work well when high level flexible reasoning is required.This paper presents an opportunistic planning-acting schema based on our model named ABC 2. The primary goal of the model is to compose pre-defined skills in an opportunistic way to achieve and intelligent behavior for robots working autonomously.