Distributed Robot Concept Storage for Speech Dialogue Integrity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for computer-aided learning of robots via voice dialog do not effectively prevent unauthorized manipulations, allowing undesirable language expressions to be learned and taught to robots, such as swear words, due to local protection mechanisms being vulnerable to manipulation.
Innovation Solution
A method that distributes concept storage across multiple robots connected via communication technology, using a consistency check based on blockchain technology to ensure that language expressions are deemed admissible only if a majority of robots agree on their admissibility, preventing unauthorized changes by requiring simultaneous manipulation across all robots.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If local protection mechanisms are used to prevent undesirable language expressions from being learned, then the robot can avoid learning inappropriate content, but the protection is vulnerable to manipulation and can be bypassed by changing the programming
Solution Approach 1:
The concept storage is segmented and distributed across multiple robots instead of being concentrated in a single robot. Each robot stores copies of the same concepts, and the system checks consistency across these distributed copies. This segmentation prevents manipulation because changing one copy is insufficient - the manipulation would be detected when comparing with other unchanged copies.
Solution Approach 2:
The system implements a feedback mechanism where each robot checks its local concepts against concepts from other robots in the network. When a concept is queried, the system retrieves corresponding concepts from multiple sources and verifies their consistency. This feedback loop detects manipulations by identifying discrepancies between distributed copies, ensuring reliability without requiring complex centralized control.
2Reliability
If concept storage is distributed across multiple robots, then protection against manipulation is enhanced, but the system complexity and communication requirements increase
Solution Approach 1:
Instead of creating a complex centralized verification system, the patent uses simple copying of concept data across multiple robots. Each robot maintains identical copies of the concept database, and verification is achieved by comparing these copies. This copying approach provides robust manipulation detection while keeping individual robot complexity low and using standard communication protocols.
3Reliability
If consistency checks are performed across multiple robots, then unauthorized manipulations are detected, but the learning process time increases
Solution Approach 1:
The system performs consistency checks on a selective basis rather than for every single learning interaction. The distributed concept storage provides inherent protection, so full consistency verification is not always necessary. This partial action approach maintains reliability while minimizing time loss during normal learning operations.
Data Source
AI summary
The invention relates to a method for computer-aided learning of a robot via a speech dialogue, wherein the robot is a predefined robot (R1) having a speech dialogue system (DS), via which a user (U) can communicate with the predefined robot (R1) verbally. In the method according to the invention a plurality of concepts (C) are examined in response to a verbal phrase (EX) that is input into the speech dialogue system (DS) by the user (U) by means of voice input within the scope of a learning dialogue and belongs to a predefined phrase category (K). The concepts (C) comprise a concept (C) stored locally in the predefined robot (R1) and one or more concepts which is/are stored in one or more further robot(s) (R2, R3, R4, R5), to which the verbal phrase (EX) input by the user (U) is transmitted, wherein a particular concept (C) of the plurality of concepts (C) describes the admissibility of verbal phrases (EX) of the predefined phrase category (K) for a verbal output by the predefined robot (R1), and the examination of a particular concept (C) delivers a first or second result. A consistency check is performed on the basis of the results of the examination of the plurality of concepts (C). In the event of an unsuccessful consistency check, the verbal phrase (EX) input by the user (U) is rejected.