AgentHack submission type
Enterprise Agents
Name
Dhananjay Mendgudli
Team name
Infinite
Team members
How many agents do you use
Multiple agents
Industry category in which use case would best fit in (Select up to 2 industries)
Manufacturing
Operations
Complexity level
Advanced
Summary (abstract)
AURA is a smart factory automation solution designed to minimize downtime and streamline incident management across diverse industrial equipment. Simulating a real-world environment with four key assets—KUKA KR 1000 Robotic Arm, Mazak CNC Lathe, RFID Smart Bins, and Rockwell Conveyor Line—AURA uses a combination of RPA bots, intelligent agents, and UiPath Apps to automate the entire lifecycle of incident detection, classification, and resolution.
Detailed problem statement
Modern manufacturing plants rely on a diverse range of automated equipment—robotic arms, CNC machines, smart inventory systems, and conveyor lines—to maintain high productivity and efficiency. However, these environments face significant challenges:
• Unplanned Downtime: Unexpected equipment failures can halt production, leading to costly delays and missed targets.
• Manual Incident Handling: Current processes for detecting, classifying, and resolving machine issues are often manual, slow, and error-prone.
• Fragmented Data: Critical information about machine health, part status, and maintenance history is scattered across multiple systems, making it difficult to respond quickly and effectively.
• Resource Bottlenecks: Maintenance teams may not have the right context or parts on hand, causing further delays and inefficiencies.
In our simulated factory scenario, featuring the KUKA KR 1000 Robotic Arm, Mazak CNC Lathe, RFID Smart Bins, and Rockwell Conveyor Line, these challenges are amplified by the complexity and interdependence of the equipment.
Detailed solution
Overview
As a two-person team, we wanted to tackle a real challenge we’ve observed in manufacturing—how to reduce downtime and speed up maintenance by automating incident detection and resolution. Our goal was to build a practical, scalable system that could be tested easily without needing live factory hardware, while leveraging UiPath’s latest agentic automation features.
- Sensor Data Simulation (Mocked API Calls)
Since we didn’t have access to live sensors, we simulated real-time sensor data using a CSV file. Each row in the CSV represents a sensor reading with the following columns:
• Timestamp
• Machine ID
• Sensor Type
• Sensor Value
• Part Name
• Part Age (days)
• Human Breach
Our RPA bot, Get Readings.xaml, reads each row sequentially and feeds it into the automation pipeline as if it were a real sensor event. This allowed us to reliably test the entire workflow end-to-end. - Issue Detection and Contextual Classification
We built an Issue Detector Agent that receives each sensor reading and decides whether it represents:
• A Non-issue (no action needed)
• A Known Issue (previously documented)
• A Maintenance Issue (requires part replacement)
• An Unknown Issue (needs human investigation)
To make these decisions accurately, the agent uses several context files:
• Threshold Matrix: Defines sensor limits per machine and part.
• Parts Details: Contains part lifecycle, vendor, and pricing info.
• Known Issue Tracker: Catalogs recurring problems and solutions.
By grounding the agent with this context, we ensured it never acts on outdated or incomplete information. The agent outputs detailed info like severity, suggested action, and issue type, which drives the next steps. - Incident Logging
Whenever the agent detects an actionable issue (anything but a non-issue), our RPA flow Create_Incident logs it into an Incident Tracking table. This table captures all relevant details such as:
• Id, Timestamp, Machine Name, Sensor Name, Reading Value
• Threshold Breached, Severity, Issue Type, Technician Assigned
• Status, Part Required, Action Taken
This centralized logging gives us full traceability and makes reporting straightforward. - Workflow Branching Based on Issue Type
a. Maintenance Issues
Our Parts Procurement Agent checks if replacement parts are needed. If the order value is below $1000, it automatically sends an order email to the vendor using Gmail integration. For orders above $1000, it triggers the Part Order Approval App to request human approval.
Once procurement is confirmed, a maintenance task is created in the Incident Task App and assigned to the appropriate technician. After the task is completed, closing comments are collected and the incident is marked closed.
b. Known Issues
The Known Issues Agent matches the problem to a known issue, assigns the right technician based on our Technician List, and provides a clear, step-by-step action plan. The task is tracked through the Incident Task App until resolution.
c. Unknown Issues
For unknown issues, a task is created in the Incident Task App for manual investigation. Once resolved, closing comments update the incident record. - Human-in-the-Loop Approvals
We use UiPath Action Center and UiPath Apps extensively to keep humans in control where needed:
• The Part Order Approval App handles procurement approvals for high-value orders.
• The Incident Task App manages technician assignments, task progress, and feedback.
This ensures automation doesn’t become a black box and that exceptions are handled efficiently. - Orchestration and Data Management
We orchestrate the entire process using UiPath Maestro, which coordinates RPA bots, agents, and human tasks seamlessly.
All decisions and actions are powered by our mapping files, which serve as the “single source of truth”:
• Threshold Matrix
• Parts Details
• Known Issue Tracker
• Technician List
• Incident Tracking
This data fabric approach ensures every part of the system is context-aware and auditable. - Closing the Loop
After any task is completed, technicians provide closing comments that update the incident record, ensuring full documentation of the resolution.
Why This Approach Works for Us
• Practical and Testable: Our CSV-based sensor simulation made it easy to develop and demo without needing physical sensors.
• Modular and Scalable: Each component—agents, RPA flows, apps—is independent, letting us iterate quickly and plan for future integration with live systems.
• Context-Aware: Grounding agents with real-time data from mapping files means decisions are always accurate and reliable.
• Human-Centric: We designed the system to keep humans involved at critical points, balancing automation with control.
• End-to-End Coverage: From detection to procurement to technician assignment and closure, the entire incident lifecycle is automated and tracked.
Demo Video
Expected impact of this automation
- Reduce equipment downtime by enabling rapid, automated detection and resolution of machine incidents.
- Accelerate response times to factory issues through context-aware agents that make real-time, data-driven decisions.
- Minimize manual effort for maintenance teams by automating incident logging, parts procurement, and technician assignment.
- Improve traceability and compliance by ensuring every action and decision is logged and auditable.
- Enhance scalability and future readiness by providing a modular system that can easily integrate with live sensors, predictive analytics, and enterprise systems.
Overall, AURA aims to transform traditional, reactive maintenance processes into a proactive, intelligent, and efficient system—delivering measurable improvements in uptime, operational efficiency, and resource utilization in smart factory environments.
UiPath products used (select up to 4 items)
UiPath Action Center
UiPath Agent Builder
UiPath Apps
UiPath Maestro
UiPath Robots
Automation Applications
Google mail
Integration with external technologies
Google Mail
Agentic solution architecture (file size up to 4 MB)
Sample inputs and outputs for solution execution
Please give the input argument in_index between 0 to 20. This will pick up one of the reading rows and show you how the reading is classified
