Modeling Unconstrained Self-Oriented Movements
Although straight point to point movements have been successfully modeled until now, many reaching movements are not straight neither in hand Cartesian nor in joint angles space, such as for example movements directed towards oneself. Addressing the generation of such free movements is fundamental to movement science, these include for example feeding movements and spontaneous gesticulation that are commonly used in everyday life and are undoubtedly an essential part of our mouvement repertoire. These reaching movements exhibit a similar bell-shaped velocity profile, but have highly curved trajectories. We argue that the curvature of path trajectories is not an undesirable side-effect of perfectly straight planned reaching movements, but is necessary and generated already at the level of the control mechanism. For example, the curvature of self-oriented movements can be attributed to fulfilling one of these two goals: (1) to avoid impossible trajectories that go through the body and (2) to avoid uncomfortable arm postures due to extreme joint angles values. One possibility is that the brain has completely separate controllers for straight and curved reaching movements and switches from one controller to the other according to the type of movement that needs to be executed. The other hypothesis is that the movement is generated by the same control mechanism, but this mechanism is adaptive in the sense that it integrates several environmental internal models such as representations of the volume and geometry of the body. By following the latter approach we have successfully modeled self-oriented reaching movements by extending a dynamical controller for generating straight point to point movements, the vector integration to endpoint (VITE) model (Bullock98,Bullock88). By introducing a body repulsive force into the model, we could reproduce the qualitative and also quantitative kinematic aspects of 3D curved movements. As our model indirectly relies on the bio-mechanical and geometrical features of the human body, it could possibly give some insights into their representation in the central nervous system for the task of reaching movements: a simple and general approximation rather than a precise and local description might be used. The VITE model is biologically plausible and brain area candidates for all the neural computations needed have been proposed. In addition the model is dynamical and thus robust to end-point and trajectory perturbations. Most importantly, some of the predictions of our model for the hand paths and their speed profiles have been validated against motion data of real human selforiented movements.