Method for representing and generalization of robot discrete track based on probability model

Abstract

The invention relates to a method for representing and generalization of a robot discrete track based on a probability model. The method comprises the following steps: teaching the discrete track, specifically, splitting the track into multiple sections and respectively teaching each section of track to obtain a data source of discrete track representation; representing the discrete track, specifically, modeling the robot track based on a plurality of GMMs, extracting a correlation relationship among the multiple sections of tracks, and representing a teaching track; and performing generalizedoutput on the track, specifically, splicing the multiple sections of track through a GMR, realizing the generalization output of the track, and making the output track have smoothness. The teaching process is simplified and the operability is high; each section of track can be smoothly spliced based on the time information; and the multi-task constraint relation of multiple mechanical arms is learned, so that the multiple mechanical arms of a robot can cooperatively finish multiple tasks.

Description

technical field [0001] The invention relates to a method for characterizing and generalizing the discrete trajectory of a robot based on a probability model. Background technique [0002] Teaching and learning can allow robots to learn how humans perform dexterous manipulation tasks in unknown environments, and generate robot trajectories that meet the needs in new environments and task objectives. The trajectory generation strategy based on teaching and learning can fully extract the characteristics of the teaching trajectory and generate robot trajectories with certain generalization. [0003] Discrete trajectories are more common in human life, such as writing Chinese characters. However, in the teaching learning field, there are few studies on feature extraction of discrete trajectories. The current research on the representation and generalization of discrete robot trajectories mainly has the following deficiencies: [0004] (1) The teaching trajectory is complex and...

AI Technical Summary

Problems solved by technology

When the learning trajectory is relatively complex, this teaching strategy is obviously cumbersome, and the teaching process requires a lot of time and cost, and it also requires certain skills to continuously design the trajectory

[0005] (2), lack of corresponding trajectory splicing strategy

Different from continuous trajectories, discrete trajectories are composed of multiple trajectories, and there are certain space constraints and time constraints between each segment of trajectories. The current research on the representation and generalization of discrete trajectories lacks corresponding strategies to integrate discrete The splicing of multi-segment trajectories cannot satisfy the relevant constraints of the original trajectories while making the final output trajectories have better smoothness.

Images