EDUARD FRANTI 1 , GHEORGHE STEFAN 2 , PAUL SCHIOPU 2 , ANCA PLAVITU 3 , TIBERIU BOROS 3 1 IMT Bucharest, Romania, edif@atlas.cpe.pub.ro 2 Politehnica University of Bucharest, Romania, gstefan@artsoc.ro 3 Research Institute for Artificial Intelligence, Romania, aplavitu@artsoc.ro, tibib@artsoc.ro, This paper presents a modular software for design artificial arms controllers. This software offers specific solutions for training the patients in order to refine the biosignals' control and optimal choice of the adequate prosthesis solutions. Using this software the patients learn how to control the biosignals by means of their visual correlation to the movements of an already attached virtual artificial arms. Based on data offered by the identification and training phases, one can choose an artificial arm solution among those available, best suited to the patient's case and abilities, along with his financial availability, required by one choice or another. Using this software the patients can save time and money, both because of delays and prosthesis mismatching elimination and because of the patient's involvement in picking up the appropriate solution. Artificial Arms, Virtual Environment, Software Asisted Training. The design of an intelligent prosthesis, for arms, involves a lot of issues that have to be solved if good results must be achieved [1]. We have to focus our efforts on the prosthesis' quality (hardware and software points of view) and patient's availability to use such prosthesis. The modeling, implementation and accommodation phases for intelligent prosthesis is obviously a complex approach, involving both manufacturers' and patient's high implications [2]. A uniquely patient's accommodation process has to be followed for every intelligent artificial arm, by means of proper initial setting-up and/or subsequent "learning". The use of biosignals requires either an analysis phase of the snag and its neighborhood (if we talk about the best case) or a complex analysis for determining as many possibilities as we can for picking up useful controlling biosignals for the artificial arm. Usually, the patient follows a series of bioelectrical tests that is going to identify the minimum necessary biosignals used in prosthesis' further control [3]. Firstly, we have to deal with the biosignals' inter- correlation. Its effects materialize as simultaneous movements for multiple joints, even if this is not what we needed. Secondly, if inter-correlation issues are solved, throughout successive testing, what follows is to verify the patients' ability to voluntarily control the distinct chosen signals. A related issue here can be seen for some patients that show difficulties in controlling muscle groups thus advancing slowly or nothing at all [4]. A third issue is due to the fact that the chosen biosignals are frequently intended for other muscle groups than those that should accomplish the desired acts. That’s why the artificial arm actuation parts must be, if possible, allocated to such biosignals that naturally command the desired movements [5]. The actuation parts that cannot be bounded up to the proper biosignals, simply because they are non-existent, require the patient's testing in order to determine his ability to control them based on other biosignals (obviously if such biosignals exist). This approach requires adequate training and generally is time consuming [2]. Besides, for some situations the results are unsatisfactory and this leads to patient's disability in fully using the acquired artificial arm' resources. It is therefore desirable that such testing can be accomplished before purchasing an artificial arm, since it is costliness. For this purpose, within our researches we've focused upon the software aided trainings, the decisions behind the biosignals' choice and upon the artificial arm Recent Researches in Automatic Control ISBN: 978-1-61804-004-6 387