This use case designs and validates an AI-enhanced IoT framework for indoor food production, aiming to reduce water consumption, optimize energy use, increase plant output, and improve process efficiency. Sensors monitor the condition of individual plants, while AI-based decision-making, edge deployment, production simulation, secure communication, and image classification support automated growing and plant monitoring. The system also addresses wireless communication security, threat detection, fallback strategies, and user trust. Its sustainability contribution lies in more localized food production, reduced resource use, fewer pesticides, optimized plant quality, and safer AI-supported food-production processes.
| # |
Technology Name |
Description |
| 1 |
PH and EC analysis |
Sensor-based PH and EC analysis is developed to decrease the daily manual work required to check PH and EC values. |
| 2 |
Edge computing architecture and relevant SW |
Time-triggered scheduling is extended to generate schedules for containers, tasks, and network communication, ensuring end-to-end timing guarantees for container orchestration. |
| 3 |
Power consumption control |
MAPE-K-based power consumption control is developed to optimize the use of devices installed in the demonstration sandbox environment. |
| 4 |
Visual plant mass estimation |
Visual plant mass estimation combines a plant-instantiation neural network with a statistical regression-based mass estimator. |
| 5 |
Communication Enhancement |
ESPAR antenna technology is used to improve wireless reliability and efficiency by dynamically switching among 4,096 radiation patterns generated through one active and 12 configurable passive elements. Combined with AI-driven optimization, the system dynamically selects effective steering vectors under changing environmental and operational conditions. |
| 6 |
ThreatGet |
ThreatGet provides an automated cybersecurity framework supporting model-based security analysis of the system architecture. It helps identify risks and define suitable security measures, including monitoring activities to mitigate cyber risks and support alignment with the EU Cyber Resilience Act. |
| KPI # |
Related Technology |
KPI Target |
| 1 |
PH and EC analysis |
Improve the efficiency ratio by 15%. |
| 2 |
Edge computing architecture and relevant SW |
Improve end-to-end latency by 10%. |
| 3 |
Power consumption control |
Improve resource usage and output by more than 10%. |
| 4 |
Visual plant mass estimation |
Improve the Incident Tolerant Time Interval by 20% and detect 75% of visual plant irregularities. |
| 5 |
Communication Enhancement |
Improve SNIR, signal-to-noise-and-interference ratio, by 75%. |
| 6 |
User awareness and trust |
Increase user awareness and trust by 10%. |
| 7 |
ThreatGet |
Increase the security level by more than 15% during the exchange of large data samples. |
| KPI # |
Baseline |
| 1 |
PH and EC parameters are checked once a day per closed system. |
| 2 |
Baseline is the measured response time with a generic Linux kernel. |
| 3 |
Baseline is the power consumption of sandbox components without AI-based control. |
| 4 |
Baseline is based on manual inspection intervals, with incidents reported during inspections. |
| 5 |
Baseline is defined as the SNIR for the omnidirectional antenna and will be determined during the final measurements. |
| 6 |
No baseline value is provided. |
| 7 |
The initial system architecture is defined without systematic cybersecurity investigation, threat and vulnerability assessment, or verification of alignment with the Cyber Resilience Act. |
| KPI # |
Validation Method |
| 1 |
Compare daily manual PH and EC checks with the number of checks triggered by the sensor-based PH and EC analysis solution. |
| 2 |
Evaluate the measured response time with the PREEMPT_RT patch in the GUT demonstrator. |
| 3 |
Compare the energy consumption of sandbox components, including the edge device, microcontrollers, and ventilation fans. |
| 4 |
Use a dedicated benchmark test dataset with manually annotated plant irregularities, and compare ML prediction accuracy and recall against the benchmark dataset. |
| 5 |
Validate communication performance through controlled measurements of the 4,096 antenna characteristics, including tests in an anechoic chamber at 1 m and 2 m radii. Use RSSI, PER, SNIR, and interference scenarios to assess wireless reliability, robustness, and continuity. |
| 6 |
Validate user awareness and trust through WP7 survey results and the AI compliance chatbot evaluation. The validation uses persona-based survey insights, structured compliance checklists, and QETAM evaluation results to assess whether supporting elements increase awareness and trust. |
| 7 |
Validate ThreatGet through model-based cybersecurity assessment and regulatory alignment checks. The assessment examines connected components for exploitable weaknesses and verifies alignment with the Cyber Resilience Act risk-based approach. |
- Industrial process phases made more effective and precise through integrated and tailored AI algorithms, methods, and tools.
- 20% improvement in prediction precision.
- 25% improvement in classification precision.
- 30% improvement in time spent on quality control.
- Improved overall productivity of the manufacturing factory by approximately 20%.
- Self-adjusted/adapted manufacturing processes above 15%.
- Providing documented methods for dynamic application of tools to improve sustainability.
- Decreased energy consumption for PoC above 10%.
The use case contributes to more effective and precise industrial process phases by combining sensor-based PH and EC analysis, edge scheduling, AI-based power control, visual plant mass estimation, communication enhancement, and cybersecurity analysis. The visual plant mass estimation technology supports prediction and classification improvements by detecting visual plant irregularities and estimating plant mass from image-based data. It also contributes directly to reducing time spent on quality control, since daily plant health inspection is currently performed manually and the proposed algorithm could reduce the time required for the daily control walk by roughly 50%, if confirmed by testing and validation. Power consumption control and resource optimization contribute to energy reduction and sustainability, while Communication Enhancement and ThreatGet improve the reliability and security of the AI-supported IoT system. Together, these technologies support safer, more efficient, and more sustainable indoor food production.
| AI Toolbox Tool |
Use within the Use Case |
| Image Classification Tool |
Used to detect and classify visual plant irregularities from image data. It supported the visual plant mass estimation component and helped automate plant-quality monitoring that was previously based on manual inspection. |
| Object detection tool |
Used to identify individual plants or plant-relevant regions in image data before visual mass estimation and irregularity detection. It supported the image-based monitoring pipeline for indoor food production. |
| Image segmentation detection with BiRefNet |
Used to segment plant regions from camera images, supporting more accurate visual plant mass estimation and plant-condition analysis. |
| Tabular Prediction Tool |
Used to analyse structured sensor and production data, including PH, EC, power consumption, environmental measurements, and resource-usage indicators. It supported AI-based decision-making for resource optimization. |
| Time-Series Anomaly Detection Tool |
Used to detect unusual temporal patterns in sensor streams, plant-condition measurements, power consumption, and communication-related operational data. It supported early identification of abnormal growing conditions or system behavior. |
| Explainability Tool |
Used to make AI-based plant monitoring, power-control, and decision-support outputs more understandable for users. It supported user trust by explaining which sensor or visual indicators contributed to AI recommendations. |
| Prompt Context Injection Tool |
Used to ground chatbot and LLM-based support functions with use-case-specific context, including system architecture, compliance requirements, cybersecurity risks, and operational constraints. |