RPA Bot Flow Correction via Neural Network Drift Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic process automation (RPA) systems face execution errors due to changes in underlying computer software and system software, leading to incorrect data entry or unexpected states, which are not recognized by the RPA processes.
Innovation Solution
The system performs automatic or user-input-based RPA flow corrections, utilizing a neural network to compare visual similarities in display images and DOM elements to determine similar RPA bot flows, allowing for the incorporation of corrective actions and user inputs to update RPA processes, and distributes these updates for use across other users after confirmation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If RPA processes are automated to perform tasks without human intervention, then productivity increases, but reliability deteriorates due to software drift causing execution errors
Solution Approach 1:
The system implements feedback mechanisms where RPA bot execution results are monitored and analyzed. When drift is detected through comparison of expected versus actual system states, the system automatically triggers correction processes. User feedback is also incorporated through confirmation requests for proposed corrections, creating a closed-loop system that continuously improves reliability while maintaining automation.
Solution Approach 2:
The system enables self-service through automated drift detection and correction generation. The RPA management system automatically identifies when software drift has occurred, generates proposed corrections by analyzing successful bot flows from other users, and implements fixes without requiring manual intervention for each incident, thereby maintaining high productivity while improving reliability.
2Reliability
If RPA bot flows are updated to adapt to software changes, then reliability improves, but device complexity increases due to correction management mechanisms
Solution Approach 1:
The system applies universality by creating a centralized correction management mechanism that serves multiple purposes: detecting drift, generating corrections, distributing updates, and validating fixes. This multi-functional approach consolidates what would otherwise be separate complex systems into a unified platform, managing reliability improvements without proportionally increasing overall complexity.
Solution Approach 2:
The system uses copying by analyzing successful bot flows from other users and replicating their correction patterns. When one user's bot flow is corrected successfully, the correction is copied and applied to other users experiencing similar drift, reducing the complexity of managing individual corrections for each user while improving reliability across the board.
3Adaptability or versatility
If user inputs are collected and distributed across multiple users, then adaptability improves, but loss of information increases due to potential errors in user-provided corrections
Solution Approach 1:
The system implements feedback loops where user-provided corrections are validated against actual system behavior and execution results. Corrections are tested in controlled environments before full deployment, and continuous monitoring provides feedback on their effectiveness. This feedback mechanism filters out inaccurate corrections while preserving valid adaptations, maintaining adaptability without sacrificing information integrity.
Solution Approach 2:
The system introduces an intermediary validation layer between user inputs and final implementation. Before user-provided corrections are deployed, they undergo automated validation, comparison with expected outcomes, and peer review through the correction management system. This intermediary process filters out erroneous corrections while preserving legitimate adaptations, reducing information loss while maintaining adaptability.
Data Source
AI summary
Robotic process automation (RPA) bot flows may be modified in the event the RPA bot flows fail to successfully complete. The RPA bot flows may be modified based on user inputs, or the RPA bot flows may be modified based on similarities to other RPA bot flows that may include similar operations. In some embodiments determinations as to similarities to other RPA bot flows are made using a neural network. In some embodiments determinations as to similarities to other RPA bot flows are made using image analysis.