Cortical network modeling for inverse kinematic computation of an anthropomorphic finger

Abstract

The performance of reaching movements to visual targets requires complex kinematic mechanisms such as redundant, multijointed, anthropomorphic actuators and thus is a difficult problem since the relationship between sensory and motor coordinates is highly nonlinear. In this article, we present a neural model able to learn the inverse kinematics of a simulated anthropomorphic robot finger (ShadowHand™ finger) having four degrees of freedom while performing 3D reaching movements. The results revealed that this neural model was able to control accurately and robustly the finger when performing single 3D reaching movements as well as more complex patterns of motion while generating kinematics comparable to those observed in human. The long term goal of this research is to design a bio-mimetic controller providing adaptive, robust and flexible control of dexterous robotic/prosthetics hands.

Cortex inspired model for inverse kinematics computation for a humanoid robotic finger

“…Furthermore, consistent with previous experimental studies, the cortical model produced slightly curved trajectories [1],[17]. Overall, the present kinematics results obtained with a physical humanoid robotic finger confirm and extend those previously obtained in simulations [13],[14]. Although the kinematics obtained both in simulation and during this robotic experiment, appear to be comparable to those observed in humans, further testing is currently in progress to directly compare these kinematics with their human counterparts while performing the same task (e.g., center-out reaching, reversal movements).…”

“…Simultaneously, the corresponding spatial displacements (Δx) of the fingertip in the 3D workspace was recorded by a motion capture system (Optotrak ® ) and then provided to the cortical model. Then, based on these spatial displacements, the cortical model estimated the joint angles (Δθ̂) that were compared to the corresponding random joint movements, providing therefore an error signal that guided the adaptation of the network parameters (e.g., w ijk , z ijkm in (3); for further details on the model implementation, see [11]–[14]).…”

“…The first one includes models that do not account for any particular neurophysiological substrate, resulting in very limited biological plausibility (e.g., [7]). The second approach proposes neural models that are biologically conceivable by incorporating particular brain structures and/or functions such as the cerebellum [8],[9] or the population vector coding processes that were previously revealed in motor/premotor areas [10]–[14]. Consistent with the second approach, recently a cortical network model able to learn the internal inverse kinematics model of a simulated anthropomorphic robot finger was proposed [13],[14].…”

Are We the Robots?

“…One fundamental problem is that the brain -as any other robotic controller aiming to command complex kinematic mechanisms-is able to learn internal models of forward and inverse sensorimotor transformations (e.g. inverse kinematic) for reaching and grasping (Gentili et al, 2011). This is a problem because the mapping between sensory and motor information is generally highly non-linear and depends on the constraints imposed by the physical features of the human or robotic hand/finger.…”

“…Inverse kinematics transforms the motion plan into joint actuator trajectories for the robot. Gentili et al (2011) propose a cortical neural model that is able to learn to control the inverse kinematics of an anthropomorphic simulated robot finger named Shadow Hand. The neural model specifically reproduces the main kinematic features of human finger movements and grip production.…”