Combining Motor Primitives for Perception Driven Target Reaching With Spiking Neurons

Abstract

Target reaching is one of the most important areas in robotics, object interaction, manipulation and grasping tasks require reaching specific targets. The authors avoid the complexity of calculating the inverse kinematics and doing motion planning, and instead use a combination of motor primitives. A bio-inspired architecture performs target reaching with a robot arm without planning. A spiking neural network represents motions in a hierarchy of motor primitives, and different correction primitives are combined using an error signal. In this article two experiments using a simulation of a robot arm are presented, one to extensively cover the working space by going to different points and returning to the start point, the other to test extreme targets and random points in sequence. Robotics applications—like target reaching—can provide benchmarking tasks and realistic scenarios for validation of neuroscience models, and also take advantage of the capabilities of spiking neural networks and the properties of neuromorphic hardware to run the models.