Continuous learning control network aiming at cerebellum dynamic sequence motion and method thereof

By introducing a Golgi cell-mediated inhibition module into the cerebellar granule cell layer, dynamically regulating the coding method of granule cells, a dynamic sequence motion learning control network was constructed, which solved the accuracy and fluency of existing cerebellar models in motor learning, and achieved efficient motor learning and execution transformation.

CN119578479AActive Publication Date: 2025-03-07TIANJIN UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411675542.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-21
Publication Date
2025-03-07
Estimated Expiration
2044-11-21

AI Technical Summary

Technical Problem

Existing cerebellar models have difficulty supporting high-level motor functions in sensorimotor learning, robust execution and flexible adjustment of motor sequences, and lack in-depth understanding of neural coding and learning mechanisms.

Method used

The feedforward and feedback inhibition module mediated by Golgi cells is introduced into the cerebellar granule cell layer to dynamically regulate the coding level and encoding mode of granule cells, thereby building a dynamic sequence motion learning control network of cerebellar.

Benefits of technology

Through the network structure of the Golgi cell inhibition module, the accuracy and fluency of the cerebellum in continuous sequence motion learning is improved, the time-direction noise interference is weakened, and efficient motion learning and execution transformation is achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119578479A_ABST
    Figure CN119578479A_ABST
Patent Text Reader

Abstract

The invention discloses a continuous learning control network aiming at cerebellum dynamic sequence motion and a method thereof. The control network comprises an input layer, a middle layer and an output layer, the middle layer is composed of a feedforward suppression module and a feedback suppression module; the input layer generates a moss fiber signal with a dynamic time sequence; the middle layer carries out sparse sampling on input moss fiber signals through granular cells, and meanwhile, the inhibition module dynamically modulates the activity level and coding mode of the granular cells; wherein the feed-forward suppression module is used for averaging a moss fiber signal discharge rate to obtain activity attributes of Golgi cells and inheriting the activity attributes to granular cells; the feedback inhibition module carries out sparse sampling treatment on the granular cells to obtain activity attributes of the Golgi cells and inherits the activity attributes to the granular cells; and the output layer sequentially outputs a result of multiplication of a connection weight between the granular cell activity range and the corresponding Purkinje cell in each time point as a cerebellum control signal. According to the invention, the problems of inaccuracy and unsmooth action conversion of a biologically constrained cerebellum network in sequence motion continuous learning are solved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Cerebellum model modeling method based on reinforcement learning

    CN107341543A

  • Redundant mechanical arm tracking control method based on echo state network

    CN113650014A

  • Target identification method and system, electronic equipment and storage medium

    CN115630675A

  • Cerebellar-like heuristic model predictive control method of redundant robot

    CN117008478A

  • Neuro-mimetic control systems and methods

    US20040075473A1