Robot Manipulator Probabilistic Workspace Applied to Robotic Assistance

Abstract

The probabilistic modelling of a robot manipulator workspace when combined with a Human-Machine Interface (HMI) allows the extraction and learning of the spatial preferences of the user. Furthermore, the knowledge of the most accessed zones of the robot’s workspace permits the bounding of the time needed for the robot to reach a given position at its workspace. From its early beginning, the use of robot manipulators within the robotic assistance field was concerned to emulate an orthopaedic arm (Fukuda et al, 2003; Zecca et al, 2002; Lopez et al, 2009). Therefore, the robot manipulator was considered as the final actuator of the assistive system where its main goal was to imitate the movements of an arm. Depending on the user/patient capabilities, the robot manipulator was commanded by either electromyographic or electro-encephalic signals (Ferreira et al, 2006a; 2006b; 2008). A robotic device controlled by a Muscle-Computer Interface (MCI) can be found in (Artemiadis & Kyriakopoulos, 2006; Lopez et al, 2007; Millan et al, 2004; Ferreira et al, 2006b; Lopez et al, 2009; Ferreira et al, 2008). In these works, the electro-miographic signal is acquired, processed, classified and converted to motion commands. The system is closed by a biofeedback loop. When used with a robot manipulator, a MCI is usually connected to a set of muscles that the patient is able to move at its own will. The number of channels used by the interface increases as increases the number of the degrees of freedom (DOF) of the robot (Yatsenko et al, 2007; Lopez et al, 2009; Ferreira et al, 2008). Thus, for a single 2DOF robot manipulator are necessary three different muscles: two to govern each DOF and a third to set a sign (if the manipulator is moving to the left or to the right), for a direct control of the robot manipulator. For robotic devices controlled by Brain-Computer Interfaces (BCI’s), the situation is analogous to the MCI; the number of signals or patterns to be extracted from the EEG (electro-encephalogram) increases as increases with the number of DOF’s –or actions– to be performed by the robot. Although most applications of BCI’s are oriented to govern a mobile robot –because of its direct analogy with a motorized wheelchair (Bastos Filho et al, 2007a; Bastos Filho, 2007; Ferreira et al, 2008; Bastos Filho et al, 2007b)– some works have been published showing the implementation of a BCI to control the movements of a robot 22