Abstract-In this article we describe the ethological inspired architecture we have developed and how it has been used to implement a humanoid goalkeeper according to the regulations of the two-legged Standard Platform League of the RoboCup Federation. We present relevant concepts borrowed from ethology that we have successfully used for generating autonomous behaviours in mobile robotics, such as the use of ethograms in robotic pets or the ideas of schemata, or the use of fixed actions patterns to implement reactivity. Then we discuss the implementation of this architecture on the Nao biped robot. Finally, we propose a method for its evaluation and validation and analyse the results obtained during RoboCup real competition, which allowed us to test first hand how it worked in a real environment.Index Terms-reactive, vision, humanoid, schema
I. ROBOTICS CONTROL ARCHITECTURES IN LITERATUREG ENERATING autonomous behaviours in mobile robotics is really a complex problem. In this section we present a non in-depth outline about those robotics control architectures close to our research. We are going to be neither exhaustive, mainly because it would be impossible to describe all control architectures in just one section, nor hierarchycal, since there would be too many criteria to take into account.Summarising the complex history of the AI, we can state that two main schools of thought have coexisted, the subsymbolic one, interested on modeling intelligence in a level similar to neurons; and the symbolic AI, which models knowledge and planning in data structures that make sense to the programmers that build them. Another way of explaining the difference between both schools is referring to their foundations: Biology in the subsymbolic AI, and cognitive psychology in the symbolic AI [9]. Hybrid systems are a pragmatic approach, where ethology based systems can be included because they successfully integrate deliberative and reactive perspectives in natural autonomous systems.Hybrid architectures intend to combine reactive and deliberative control, and usually consist of three components: a reactive layer, a planner, and a layer that links the other two. Opportunistic architectures are a subset of the hybrid architectures that take its name from Barbara Hayes-Roth approach to hybrid control [7]. The agents architecture on her system was also made up by three components: an event-triggered reactive level, an strategic planner, and a control process in charge of matching triggered actions with the generated plan. A similar architecture is used in where the term "agent" is used to name each of the three modules of the system.Another implementation of these ideas are RAPs (Reactive Action Packages) proposed by Firby [5]. RAPs were designed to allow the reactive execution of symbolic plans. In this way, each RAP defines different alternatives of execution depending on the environment, and an agenda is used to select the next action to execute. Another approach is the TCA (Task Control Architecture) by Simmons [17], which inte...