Robot Movement Embedding for Anomalous State Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robots face challenges in consistently performing tasks due to difficulty in training, potential for errors leading to damage, and occurrence of anomalous states such as singularity, which can hinder task completion and cause accidents.
Innovation Solution
A computing system employs an anomaly detection model to classify robot states as anomalous by encoding sensor data into embeddings and applying a specification to detect deviations from expected behaviors, preventing further actions when anomalies are detected.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If reinforcement learning is used to train robots to complete tasks through trial-and-error, then robot task completion capability is improved, but training time and computational resources increase significantly
Solution Approach 1:
The patent pre-computes and stores cost maps during an offline training phase, capturing the robot's cost-to-go information for different states and actions. This preliminary action allows the robot to make rapid decisions during online operation without performing real-time trial-and-error learning, thus resolving the contradiction between task completion capability and training time.
2Speed
If robots perform actions based on learned policies, then task execution speed is improved, but risk of errors and damage increases
Solution Approach 1:
The patent implements a feedback mechanism where the robot continuously monitors its current state and compares it against pre-computed cost maps. The cost maps provide feedback information about the expected costs of different actions, allowing the robot to select actions that minimize risk while maintaining execution speed. This feedback loop resolves the contradiction between speed and reliability.
Solution Approach 2:
The patent pre-computes cost maps that incorporate risk and cost information for various actions before the robot executes tasks. This beforehand cushioning provides a safety buffer by having pre-analyzed action costs available during execution, allowing the robot to avoid high-risk actions without slowing down real-time operation, thus resolving the contradiction between execution speed and error prevention.
3Device complexity
If the robot operates without anomaly detection, then system complexity is reduced, but risk of accidents and damage increases
Solution Approach 1:
The patent pre-computes cost maps during an offline phase that encode safety and cost information for various states and actions. This preliminary action embeds anomaly detection capabilities into the pre-computed maps, allowing the system to detect anomalous states during operation without adding complex real-time detection algorithms, thus resolving the contradiction between system complexity and accident risk.
Data Source
AI summary
Provided is a process, including: obtaining, with a computer system, access to a specification indicating which regions of an embedding space are designated as anomalous relative to vectors in the embedding space characterizing past behavior of a first instance of a dynamical system; receiving, with the computer system, multi-channel input indicative of a state of a second instance of the dynamical system; and classifying, with the computer system, whether the state of the second instance of the dynamical system is anomalous by: encoding the multi-channel input into a vector in the embedding space; causing the specification to be applied to the vector; obtaining a result of applying the specification to the vector; and classifying whether the state of the second instance of the dynamical system is anomalous based on the result; and storing the classification in memory.