The latest research topics being explored in the SOC2 lab.
CURRENT
RESEARCH
AREA
Sustainable Computing on the Edge–Cloud Continuum
AI services are becoming central to modern life, from healthcare to urban analytics, but they come at a high cost: rising energy use and a growing carbon footprint. Making computing more powerful and environmentally responsible is a pressing challenge. Our research develops carbon-aware orchestration mechanisms that manage where and how services run across the edge–cloud continuum. We design workload distribution strategies and scheduling algorithms that adapt to dynamic conditions, balancing performance, energy efficiency, and emissions. This work spans multiple paradigms, from virtual machines and containers to serverless computing, and focuses on orchestrating services through placement, migration, and resource provisioning. A key direction is integrating renewable energy into orchestration decisions, where supply is variable and workloads are unpredictable. By embedding sustainability into the core of edge–cloud systems, we enable digital infrastructures that reduce CO₂ emissions and operational costs while supporting scalable AI services for society.
LLM@Edge and Multi-Cloud Optimization
Large Language Models (LLMs) are transforming AI, but their deployment is often limited by high energy costs, latency, and the vast resources they require. This research area explores how to make LLMs feasible and sustainable across the edge–cloud continuum, enabling advanced intelligence closer to users while reducing environmental impact. Our work develops orchestration and scheduling mechanisms that partition LLMs across heterogeneous infrastructures. Lightweight components run on edge devices for immediate responsiveness, while complex tasks are offloaded to cloud or high-capacity nodes. Scheduling decisions adapt to device capabilities, network conditions, workload demands, and energy-aware metrics, with a focus on lowering the carbon footprint through renewable-powered and energy-efficient resources. By integrating distributed systems principles with advanced orchestration frameworks, this research unlocks scalable, low-latency, and sustainable LLM services. Applications include smart cities, healthcare, smart homes, and IoT ecosystems, where LLM@Edge enables real-time, resource-sensitive, and context-aware AI.
Previous Research Areas
Research topics and outputs that members have participated in.
Service Orchestration on Integrated Heterogeneous Networking Infrastructure
Reliable access to digital services becomes critical in situations where conventional networks are limited, disrupted, or unavailable. This research line explores service orchestration in integrated heterogeneous networking infrastructures, with a focus on the Space–Air–Ground Integrated Network (SAGIN). By combining satellites, aerial platforms, and terrestrial networks, SAGIN enables resilient global connectivity but introduces challenges in service routing, task scheduling, and service placement. Our work develops mechanisms to orchestrate services seamlessly across these layers, ensuring reliability, low latency, and adaptability under highly dynamic conditions. These advances are particularly relevant for pre- and post-disaster management (e.g., flooding, wildfires), where resilient cloud-edge services can support emergency response and protect communities. Beyond disaster scenarios, this research also opens opportunities in remote communications, large-scale IoT integration, and global monitoring. This line of work is pursued in collaboration with Trinity College Dublin (TCD) and Munster Technological University (MTU), strengthening interdisciplinary efforts toward resilient and adaptive digital infrastructures.
The submitted manuscripts are under review.
Connecting Communities to Smart Urban Environments through the Internet of Things
This research line was part of the ENABLE Research Program at Trinity College Dublin, which sought to connect communities to smarter, more sustainable, and more inclusive urban environments through the Internet of Things. A key challenge in this vision is ensuring that smart city services run efficiently and adapt to the needs of diverse citizens, all while operating at scale across complex infrastructures. Our work addressed this challenge by advancing multi-access edge computing (MEC) as the foundation for delivering low-latency and resource-efficient services close to where data is generated. We developed a reinforcement learning–based service placement method that dynamically cached application services on edge servers, reducing backhaul traffic and improving responsiveness. To support real-world deployment, we proposed approaches for IoT interoperability across heterogeneous systems and designed a gossip-based dissemination mechanism to enable context-aware information exchange among autonomous vehicles. Recognizing the social dimension of urban innovation, we also contributed to age-friendly traffic management, introducing methods to smooth speed variability and make transportation systems more accessible and inclusive.
Publications:
- Bailey, J.M., Tabatabaee Malazi, H., & Clarke, S. (2021). Smoothing Speed Variability in Age-Friendly Urban Traffic Management. In Intl. Conf. on Computational Science (ICCS). Springer.
- Rasool Chaudhry, S., Tabatabaee Malazi, H., Dhara, S., Kazmi, A., & Clarke, S. (2023). Towards Context-aware Information Dissemination in Autonomous Networks of Vehicles. IEEE Communications Standards Magazine, 7(2): 8-15.
- Kazmi, A., Rasool Chaudhry, S., Tabatabaee Malazi, H., Dhara, S., & Clarke, S. (2022). Interoperable Internet of Things for Smart Transportation Systems in Circular Cities. Computer, 55(12): 86-97.
- Tabatabaee Malazi, H., Chaudhry, S.R., Kazmi, A., Palade, A., Cabrera, C., White, G., & Clarke, S. (2022). Dynamic Service Placement in Multi-access Edge Computing: a Systematic Literature Review. IEEE Access, 10: 143150-143165.
- Tabatabaee Malazi, H., & Clarke, S. (2021). Distributed Service Placement and Workload Orchestration in a Multi-access Edge Computing Environment. In IEEE Intl. Conf. on Services Computing (SCC).
Complex Event Processing in Smart City Edge Computing
Air pollution and traffic congestion are among the most pressing challenges in modern cities, where timely responses depend on making sense of massive sensor data streams. In this research line, we developed methods to detect and localize complex events such as pollution spikes in real time, enabling faster and more reliable decision-making for urban management. Our approach combined a publish–subscribe architecture with software-defined networking (SDN) to improve scalability, and we extended the T-Rex event processing language with spatial operators to support geospatial reasoning. These innovations enabled efficient, distributed analysis of environmental and IoT data at the edge of the network, supporting applications in pollution control, smart transportation, and urban resilience.
Publications:
- Khazael, B., Vahidi-asl, M., & Tabatabaee Malazi, H. (2023). Geospatial Complex Event Processing in Smart City Applications. Simulation Modelling Practice and Theory, 122: 102675. Elsevier.
- Kazmi, A., Rasool Chaudhry, S., Tabatabaee Malazi, H., Dhara, S., & Clarke, S. (2022). Interoperable Internet of Things for Smart Transportation Systems in Circular Cities. Computer, 55(12): 86-97. IEEE.
- Khazael, B., & Tabatabaee Malazi, H. (2020). Distributed coordination protocol for event data exchange in IoT monitoring applications. In 11th Intl. Conf. on Information and Knowledge Technology (IKT2020).
Sensor-Based Human Activity Recognition
Human activity recognition is a cornerstone for applications such as mobile health monitoring, assisted living, and sports analytics, where understanding daily behavior enables smarter and safer environments. In this research line, we investigated activity recognition using smart home sensors, smartphones, and wearable devices like smartwatches. Our work contributed in three ways: building multi-sensor datasets for realistic activity tracking, developing methods for fine-grained motion recognition (e.g., detecting tennis strokes), and improving the accuracy of complex activity recognition through emerging pattern mining, machine learning, and sensor fusion. These contributions advanced the reliability of sensor-based activity recognition and opened pathways for applying it in real-world healthcare and sport contexts.
Publications:
- Taghavi, S., Davari, F., Tabatabaee Malazi, H., & Abin, A. A. (2019). Tennis stroke detection using inertial data of a smartwatch. In 9th Intl. Conf. on Computer and Knowledge Engineering (ICCKE) (pp. 466-474). IEEE. (Add the link to the dataset)
- Tabatabaee Malazi, H., & Davari, M. (2018). Combining Emerging Patterns with Random Forest for Complex Activity Recognition in SmartHomes. Applied Intelligence, 48(2): 315-330. Springer.
- Esfahani, P., & Tabatabaee Malazi, H. (2017). PAMS: A new position-aware multi-sensor dataset for human activity recognition using smartphones. In 19th Intl. Symp. on Computer Architecture and Digital Systems (CADS’17). IEEE.
Social Sensing Systems
Social media platforms generate massive volumes of real-time data that reflect how people collectively report, interpret, and react to events. In this research line, we treated humans as social sensors and developed methods to detect and localize events by mining and processing large-scale data streams. Using data mining, machine learning, and stream processing, we advanced techniques for reliable event extraction, credibility assessment, and fine-grained event localization. Our work demonstrated how social sensing can support smart cities, crisis management, and other time-critical applications, showing the potential of human-driven data to complement traditional sensing systems.
Publications:
- Shahraki Khodabandeh, Z., Fatemi, A., & Tabatabaee Malazi, H. (2019). Evidential fine-grained event localization using Twitter. Information Processing & Management, 56(6): 102045. Elsevier.
- Toosinezhad, Z., Mohamadpoor, M., & Tabatabaee Malazi, H. (2019). Dynamic windowing mechanism to combine sentiment and N-gram analysis in detecting events from social media. Knowledge and Information Systems, 60(1): 179-196. Springer.
- Taghizadeh Naderi, P., Tabatabaee Malazi, H., Ghassemian, M., & Haddadi, H. (2016). Quality of claim metrics in social sensing systems: a case study on IranDeal. In 6th Intl. Conf. on Computer and Knowledge Engineering (ICCKE) (pp. 129-135). IEEE.
