Abstract: A transition method of locomotion gait of a robot includes: executing a deployment procedure multiple times, each execution includes: randomly selecting a source policy and a destination policy, simulating a transition operation from the source policy to the destination policy, and recording a transition configuration and a transition result to a transition database, where each policy is a neural network model, and a latent state in the transition configuration is a hidden layer of the neural network model of the source policy. The method further includes: training a transition-net according to the transition database, and performing the following steps by a meta-controller disposed on the robot: selecting two gait policies as an active policy and a queued policy, executing the active policy, inputting the two policies to the transition-net to obtain a success probability, and when the success probability is greater than a threshold, executing the queued policy.

Abstract: A transition method of locomotion gait of a robot includes: executing a deployment procedure multiple times, each execution includes: randomly selecting a source policy and a destination policy, simulating a transition operation from the source policy to the destination policy, and recording a transition configuration and a transition result to a transition database, where each policy is a neural network model, and a latent state in the transition configuration is a hidden layer of the neural network model of the source policy. The method further includes: training a transition-net according to the transition database, and performing the following steps by a meta-controller disposed on the robot: selecting two gait policies as an active policy and a queued policy, executing the active policy, inputting the two policies to the transition-net to obtain a success probability, and when the success probability is greater than a threshold, executing the queued policy.

Abstract: A method for automatically classifying transition motion includes following steps performed by a computing device: obtaining a plurality of transition motions, with each transition motion being associated with a source motion, a destination motion, and a transition mechanism converting the source motion into the destination motion; extracting a property vector from each transition motion and thereby generating a plurality of property vectors, wherein each property vector includes a plurality of transition properties; and performing a clustering algorithm according to the property vectors to generate a plurality of transition types.

Abstract: An image data augmentation device is provided, which comprises an image capturing circuit and a processor. The processor is configured to execute the following operations: using an object detection model to generate multiple bounding boxes and respective classification labels of the bounding boxes from an image; identifying an overlap ratio between the two bounding boxes, and determining whether the overlap ratio between the two bounding boxes is more than a ratio threshold, where the two bounding boxes have the same classification label; and when the overlap ratio between the two bounding boxes is more than the ratio threshold, deleting one of the two bounding boxes to update the bounding boxes, thereby by using the bounding boxes and the respective classification labels of the bounding boxes for executing machine learning.

Abstract: A method for automatically classifying transition motion includes following steps performed by a computing device: obtaining a plurality of transition motions, with each transition motion being associated with a source motion, a destination motion, and a transition mechanism converting the source motion into the destination motion; extracting a property vector from each transition motion and thereby generating a plurality of property vectors, wherein each property vector includes a plurality of transition properties; and performing a clustering algorithm according to the property vectors to generate a plurality of transition types.

Abstract: A method for establishing a complex motion controller includes following steps: obtaining a source controller and a destination controller, wherein the source controller is configured to generate a source motion, and the destination controller is configured to generate a destination motion; determining a transition tensor between the source controller and the destination controller, wherein the transition tensor has a plurality of indices, one of the plurality of indices corresponds to a plurality of phases of the source motion; calculating a plurality of transition outcomes of the transition tensor and recording the plurality of transition outcomes according to the plurality of indices; calculating a plurality of transition qualities according to the plurality of transition outcomes; and searching for an optimal transition quality from the plurality of transition qualities to establish a complex motion controller for generating a complex motion corresponding to one of the plurality of phases.

Abstract: A method for establishing a complex motion controller includes following steps: obtaining a source controller and a destination controller, wherein the source controller is configured to generate a source motion, and the destination controller is configured to generate a destination motion; determining a transition tensor between the source controller and the destination controller, wherein the transition tensor has a plurality of indices, one of the plurality of indices corresponds to a plurality of phases of the source motion; calculating a plurality of transition outcomes of the transition tensor and recording the plurality of transition outcomes according to the plurality of indices; calculating a plurality of transition qualities according to the plurality of transition outcomes; and searching for an optimal transition quality from the plurality of transition qualities to establish a complex motion controller for generating a complex motion corresponding to one of the plurality of phases.

Abstract: A method for training a locomotion controller of a robotic animal comprising obtaining a motion data of a reference animal, an environmental parameter, and a disturbance parameter; generating a plurality of primitive distributions and a first primitive influence according to the motion data by a policy network; selecting a current state of the reference animal from the motion data and set an adapting state of the reference animal; generating a second primitive influence, by the policy network, according to the current state, the adapting state, and the plurality of primitive distributions at least; and training the policy network according to a result determined by a discriminator according to the first primitive influence and the second primitive influence.

Abstract: A hyper-parameter configuration method of time-series forecasting model comprises storing N datasets respectively corresponding to N products; determining a forecasting model; and performing a hyper-parameter searching procedure. The hyper-parameter searching procedure comprises generating M sets of hyper-parameters; applying each set of hyper-parameters to the forecasting model; training and validating the forecasting model respectively according to two strategies to generate two error arrays, wherein the two strategies selects the training dataset and the validation dataset from N datasets in different two data dimensions, performing a weighting computation or an ordering operation according to two weights and the two error arrays and searching for a target set of hyper-parameters, wherein two error values corresponding to the target set of hyper-parameters in the two error arrays are two relative minimums.

Abstract: A signal quality index evaluation circuit, comprises: a surrounding sensor; a zero-phase filter; and an evaluation circuit. The surrounding sensor senses its surrounding to generate a reference correction signal. The zero-phase filter is configured to generate a clean biological signal according to a biological signal and the reference correction signal, wherein the clean biological signal includes a plurality of period signals, and each one of the period signals has a biological value. The evaluation circuit is configured to calculate norm range according to the clean biological signal and one or more of the biological values of the period signals, and determine a difference between each one of the biological values corresponding to each one of the period signals and the norm range, the evaluation circuit is further configured to calculate and output a signal quality index according to the differences.

Abstract: A method for training a locomotion controller of a robotic animal comprising obtaining a motion data of a reference animal, an environmental parameter, and a disturbance parameter; generating a plurality of primitive distributions and a first primitive influence according to the motion data by a policy network; selecting a current state of the reference animal from the motion data and set an adapting state of the reference animal; generating a second primitive influence, by the policy network, according to the current state, the adapting state, and the plurality of primitive distributions at least; and training the policy network according to a result determined by a discriminator according to the first primitive influence and the second primitive influence.

Abstract: A method for dynamically switching blood pressure measurement model, adapted to a wearable blood pressure measurement device with a biosignal sensing assembly and a processor, wherein said biosignal sensing assembly comprises two exposed electrodes, comprises: obtaining potential difference by said two exposed electrodes; determining whether potential difference is smaller than potential threshold by processor; obtaining first biosignal of specified user by biosignal sensing assembly when potential difference is smaller than the potential threshold; calculating first blood pressure value by processor according to at least first biosignal and first blood pressure model and outputting first blood pressure value; obtaining second biosignal of specified user by biosignal sensing assembly when potential difference is not smaller than the potential threshold, wherein types of first and second biosignal are different; and calculating second blood pressure value by processor according to second biosignal and second b

Abstract: A method of establishing blood pressure model comprising: obtaining a plurality of first physiologic data of a plurality of general users and a plurality of first blood pressure data of the plurality of general users; performing a deep learning algorithm to establish a general blood pressure model according to the plurality of first physiologic data and the plurality of first blood pressure data, wherein the general blood pressure model has a parameter set and a loss function; obtaining a second physiologic data of a specific user and a second blood pressure data of the specific user; generating a blood pressure estimation according to the second physiologic data and the parameter set; calculating an error according to the blood pressure estimation, the second blood pressure data and the loss function; and adjusting the parameter set to establish a specific blood pressure model according to the error.

Abstract: A product order prediction method adapts to a production planning system. The method comprises obtaining a reference data related to a next reference order and a current actual order, performing an algorithm based on the neural network model according to the reference data and the current actual order to generate a feature vector, and performing another algorithm based on the neural network model according to the feature vector to output a next predicted order to the production planning system, for the production planning system to generate an operation plan of a production line of the product according to the next predicted order.