Automatic neural-based pattern classification of motion behaviors in autonomous robots Abraham Prieto, Francisco Bellas n , Pilar Caaman ˜o, Richard J. Duro Grupo Integrado de Ingenierı ´a, Universidade da Corun ˜a, Spain article info Available online 3 August 2011 Keywords: Hybrid intelligent systems ANN for classification Motion pattern analysis Behavior detection Cognitive developmental robotics abstract This paper addresses the problem of providing autonomous robots with a system that allows them to classify the motion behavior patterns of groups of robots present in their surroundings. It is a first step in the development of a cognitive model that can detect and understand the events occurring in the environment that are not due to the robot’s own actions. The recognition of motion patterns must be achieved from the input data acquired by the robot through its camera during real time operation and, consequently, it can be addressed as a high dimensional dynamic pattern classification problem. Artificial Neural Networks (ANN) have been widely used in this type of classification problems, where a preprocessing stage is typically introduced in order to reduce dimensionality. In this stage, the processing window size and the dimensional transformation parameters must be selected according to specific domain knowledge, and they remain fixed during the ANN classification process. Such an approach is not applicable here as there is no prior information on the number of robots present or the dimensional reduction level required to describe the possible robot motion behaviors. Consequently, this work proposes a hybrid approach based on the application of a classification system called ANPAC (Automatic Neural-based Pattern Classifier) that uses a variable size ANN to perform the classification and an advisor module to adjust the preprocessing parameters and, consequently, the size of the ANN, depending on the learning results of the network. The components and operation of ANPAC are described in depth and illustrated using an example related to the recognition of behavior patterns in the motion of flocks. & 2011 Elsevier B.V. All rights reserved. 1. Introduction Autonomous robots that operate completely isolated from the actions of other robots and their consequences, assuming that their environment will be modified only by their own actions, have been studied for decades [1,2]. However, such assumptions are not realistic if fully autonomous robots performing their tasks taking into account that there are other inhabitants in the world where they are placed are to be obtained. That is, robotic architectures must include concepts like attention, learning by observation, behavior detection, empathy and so on, to really take into account what is happening in the environment and thus perform a correct selection of actions [3,4] and this is usually the realm of cognitive architectures. Cognitive developmental robotics (CDR) is a novel research field that deals with the design and implementation of cognitive architectures for autonomous robots [5]. Specifically, the main objective of this approach is to obtain open-ended autonomous learning systems that continually adapt to their environment, as opposed to the classical approach of constructing robots that carry out particular, predefined tasks. The basic concept behind CDR is ‘‘physical embodiment’’, that allows structuring information through interactions with the environment and other robots. CDR uses a developmental model that starts from the fetal sensorimotor map- ping in the womb and moves up to social behavior learning through body representation, motor skill development, and spatial percep- tion. One of the key aspects of CDR is the autonomous development of social behaviors such as early communications, action execution and understanding, vocal imitation, joint attention, and empathy development. These topics are not specific to the area of CDR and designing social machines has been a very active topic in autono- mous systems and related fields [6,7,3]. However, in CDR the social aspects are studied from a quite different point of view [5,8]. Thus, when the aim is to create fully autonomous developmental robots, one of the highest-level concepts that must be studied is that of consciousness [9]. Conscious robots must be capable of analyzing the consequences of their actions and distinguish them from those of the actions applied by others. This implies that, the cognitive model in Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/neucom Neurocomputing 0925-2312/$ - see front matter & 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.neucom.2011.02.027 n Correspondence to: Universidade da Corun ˜ a, Escuela Polite ´ cnica Superior, Mendiza ´ bal s/n, 15403 Ferrol (A Corun ˜ a), Spain. Tel.: þ34 981 337400x3886; fax: þ34 981 337410. E-mail addresses: abprieto@udc.es (A. Prieto), fran@udc.es (F. Bellas), pcsobrino@udc.es (P. Caaman ˜ o), richard@udc.es (R.J. Duro). Neurocomputing 75 (2012) 146–155