Neuromorphic Eye-in-Hand Visual Servoing

Robotic vision plays a major role in factory automation to service robot applications. However, the traditional use of frame-based camera sets limitation on continuous visual feedback due their low sampling rate and redundant data real-time image processing, especially case high-speed tasks. Event cameras give human-like capabilities such as observing dynamic changes asynchronously at high temporal resolution ($1\mu s$) with latency wide range. In this paper, we present servoing method using an event switching control strategy explore, reach grasp achieve manipulation task. We devise three surface layers active events directly process stream from relative motion. A purely based approach is adopted extract corner features, localize them robustly heat maps generate virtual features for tracking alignment. Based feedback, motion controlled make upcoming converge desired spatio-temporal space. The controller switches its sequence operation establish stable grasp. (EVBS) validated experimentally commercial manipulator eye-in-hand configuration. Experiments prove effectiveness EBVS track objects different shapes without need re-tuning.