Robot Skill Sequencing With Learned Branching Conditions

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing robotic systems struggle to efficiently sequence and execute manipulation skills in dynamic environments, particularly when tasks involve branching conditions that require manual definition and sensor data processing.

Innovation Solution

A method for controlling robots that involves training hidden semi-Markov models from demonstrations, creating precondition and final condition models, and generating a composed robot trajectory model to automatically sequence skills in sequence or branches, eliminating the need for manual definition of branching conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If manual definition of branching conditions is used to sequence skills, then the robot can execute tasks with different skill sequences, but the system complexity increases and robustness decreases due to incomplete state space coverage

Engineering Contradiction:

Improveskill sequencing capabilityVSAvoidbranching condition definition complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system automatically determines skill sequencing and branching conditions through learned transition models rather than requiring manual programming. The robot serves itself by autonomously selecting appropriate skills based on learned state transitions from demonstrations, eliminating the need for operators to manually define branching logic and state space conditions.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces manual mechanical programming of branching conditions with a data-driven learning system. Instead of manually configuring state-space coverage and transition rules, the system learns skill transitions automatically from demonstrated tasks, substituting the manual configuration process with an automated learning mechanism that processes demonstration data to infer appropriate skill sequences.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If manual sensor data processing is used to evaluate branching conditions, then the robot can make decisions based on current state, but the time consumption increases and robustness decreases

Engineering Contradiction:

Improvetask execution automationVSAvoidbranching condition evaluation time

Core Design Contradiction:

Ease of operationVSLoss of time

Solution Approach 1:

The system performs preliminary learning of skill transitions during an offline phase using demonstration data. By pre-processing and learning the transition models beforehand, the system avoids time-consuming sensor data processing and branching condition evaluation during actual task execution. The pre-learned models enable rapid skill selection without real-time manual intervention.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The robot autonomously determines skill transitions using pre-learned models without requiring manual sensor data processing during execution. The system serves itself by automatically selecting skills based on the current state and learned transition probabilities, eliminating the need for operators to process sensor data and evaluate branching conditions in real-time.

Inventive Principle:

Principle #25Self-service

3Productivity

If manual definition of branching conditions is required, then the task can be executed with specific skill sequences, but the effort required increases and the system becomes less robust to unanticipated situations

Engineering Contradiction:

Improvetask execution efficiencyVSAvoidtask setup effort

Core Design Contradiction:

ProductivityVSEase of manufacture

Solution Approach 1:

The patent replaces manual task setup and branching condition definition with an automated learning system. Instead of requiring operators to manually configure task-specific branching logic, the system learns skill transitions automatically from demonstrations and generates executable task plans without manual intervention, significantly reducing setup effort while maintaining execution efficiency.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs the labor-intensive work of learning skill transitions and determining branching conditions in advance during an offline training phase. This preliminary learning action transfers the computational burden from runtime decision-making to offline model training, enabling efficient task execution with minimal manual effort during deployment.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12208522B2Method for controlling a robot and robot controller

Publication Date: 2025.01.28 ROBERT BOSCH GMBH

AI summary

A method for controlling a robot. The method includes providing demonstrations for performing each of a plurality of skills; training from the demonstrations, a robot trajectory model for each skill, each trajectory model is a hidden semi-Markov model having one or more initial states and one or more final states; training, from the demonstrations, a precondition model for each skill comprising, for each initial state, a probability distribution of robot configurations before executing the skill, and a final condition model for each skill comprising, for each final state, a probability distribution of robot configurations after executing the skill; receiving a description of a task, the task includes performing the skills of the plurality of skills in sequence and/or branches; generating a composed robot trajectory model; and controlling the robot according to the composed robot trajectory model to execute the task.