# Iman Esfandiyar

> Robotics and AI researcher based in Poznań, Poland, working on dense 3D mapping, traversability estimation, motion planning and control for autonomous mobile robots in agriculture and forestry. PhD student in robotics at Poznan University of Technology and Chief Specialist in Robotics & AI at Łukasiewicz Research Network – Poznań Institute of Technology.

This site publishes the full text of Iman Esfandiyar's journal articles as HTML and Markdown, with complete citation data (BibTeX, RIS) and DOIs, so they can be read, summarised and cited accurately. When citing, please cite the published version via its DOI.

- ORCID: https://orcid.org/0000-0002-3360-7728
- Google Scholar: https://scholar.google.com/citations?user=YvZgBG4AAAAJ
- Email: iman@esfandiyar.com
- LinkedIn: https://www.linkedin.com/in/imanesfandiyar/
- Home page: https://esfandiyar.com/

## Hire or work with Iman Esfandiyar

Robotics and AI engineer based in Poznań, Poland. Available for freelance and contract programming, consulting and R&D projects in robotics and AI, on-site in Poznań, remotely, or on-site elsewhere in Poland and the EU; open to job offers. Works with clients in English or Polish.

- [Profile: skills, services, projects and experience](https://esfandiyar.com/profile.md): C++, Python, ROS / ROS 2, neural networks, reinforcement learning, sensor fusion, localization, motion planning, computer vision, simulation environments; autonomous mobile robots, field and agricultural robots, driver assistance, drone swarms.
- Web version of the profile: https://esfandiyar.com/profile/

## Publications with full text

- [Dual-Link Data Resilient Edge-to-cloud Communication Framework for Agricultural Robots](https://esfandiyar.com/publications/dual-link-edge-to-cloud-agricultural-robots/paper.md): Iman Esfandiyar and Kamil Młodzikowski, in IoTAI 2025: The Second International Conference on IoT-AI, pp. 27-34, 2025. doi:10.5281/zenodo.17735730. Preferred citation: I. Esfandiyar and K. Młodzikowski, "Dual-Link Data Resilient Edge-to-cloud Communication Framework for Agricultural Robots," in Proc. IoTAI 2025: The Second International Conference on IoT-AI, IARIA, 2025, pp. 27-34. ISBN: 978-1-68558-286-9.
  - Web page: https://esfandiyar.com/publications/dual-link-edge-to-cloud-agricultural-robots/ · BibTeX: https://esfandiyar.com/publications/dual-link-edge-to-cloud-agricultural-robots/cite.bib · RIS: https://esfandiyar.com/publications/dual-link-edge-to-cloud-agricultural-robots/cite.ris
  - Summary: Problem: field robots in rural farmland need to exchange data with cloud services, but high-bandwidth connectivity there is patchy and links drop as the robot moves away from the gateway. Approach: a dual-link framework in which long-range LoRa (LPWAN) carries essential control and monitoring messages, while Wi-Fi carries bulk sensor and image data over the Zenoh protocol, built on ROS 2. Data router: switches the robot between Transfer Mode (forward data over Wi-Fi) and Storage Mode (record locally in ROS 2 bags) so no data is lost when Wi-Fi degrades. Link-failure prediction: an onboard XGBoost-based anomaly detector monitors heartbeat timing and flags Wi-Fi degradation before an outage, triggering the router. Field results: Wi-Fi latency stayed under 100 ms within 240 m of the gateway; LoRa stayed reliable beyond 350 m at about 0.1 s latency; the router averaged 0.8 s overlap when entering Storage Mode. Scope and future work: designed to scale to robot swarms via ROS 2 namespaces, Zenoh and LoRa multicast (not yet validated on a fleet); developed as a use case for the EU Horizon ICOS meta operating system.
- [Towards a Functional Continuum Operating System – ICOS MetaOS](https://esfandiyar.com/publications/icos-metaos-functional-continuum-operating-system/paper.md): Artur Jaworski et al., in Global Internet of Things and Edge Computing Summit, GIECS 2024, Communications in Computer and Information Science, vol. 2328, pp. 119-133, 2025. doi:10.1007/978-3-031-78572-6_8. Preferred citation: Jaworski, A., Kotlinski, M., Zrazinska, I., Masip-Bruin, X., Garcia, J., Esfandiyar, I. (2025). Towards a Functional Continuum Operating System – ICOS MetaOS. In: Presser, M., Skarmeta, A., Krco, S., González Vidal, A. (eds) Global Internet of Things and Edge Computing Summit. GIECS 2024. Communications in Computer and Information Science, vol 2328. Springer, Cham. https://doi.org/10.1007/978-3-031-78572-6_8
  - Web page: https://esfandiyar.com/publications/icos-metaos-functional-continuum-operating-system/ · BibTeX: https://esfandiyar.com/publications/icos-metaos-functional-continuum-operating-system/cite.bib · RIS: https://esfandiyar.com/publications/icos-metaos-functional-continuum-operating-system/cite.ris
  - Summary: Context: ICOS is an EU Horizon project building a meta-operating system (metaOS) for the cloud–edge–IoT continuum, aimed at device heterogeneity, scalable and secure application execution, lower integration costs and less cloud-provider lock-in. Architecture: distributed ICOS Controllers, each with a Meta-Kernel layer (continuum and runtime management, telemetry), an Intelligence layer (predictive monitoring, federated learning) and a Security layer, manage ICOS Agents that run on every node, from clusters to constrained edge devices. Use Case 1, agriculture: the AgroRob autonomous field robot (GNSS, RGB and depth cameras, ROS 2, YOLOv8) detects and localizes weeds and crops to map the field; ICOS is to keep the robot operating in rural areas with low network coverage and high latency. Use Case 2, railway: Worldsensing's Railway Structural Alert Monitoring on the FGC Lleida–La Pobla line, with LoRaWAN tiltmeters, extensometers and piezometers on a 4 km stretch, an edge IoT gateway on 4G, and Google Cloud. Alpha-release integration: both pilots onboarded edge nodes with Nuvla and deployed the ICOS Telemetry and VPN agents; on the Worldsensing gateway, Docker was chosen over Kubernetes because of lower CPU use. Next steps: translate Docker Compose and Kubernetes manifests into ICOS application manifests for the Beta release, then use the Intelligence layer for predictive maintenance and better image-recognition models.
- [Mapping the Unseen Ground: A Survey on Dense Mapping and Traversability Methods for Agricultural and Forestry Robots](https://esfandiyar.com/publications/mapping-the-unseen-ground/paper.md): Iman Esfandiyar and Dominik Belter, IEEE Access 13, 204285-204314, 2025. doi:10.1109/ACCESS.2025.3638877. Preferred citation: I. Esfandiyar and D. Belter, "Mapping the Unseen Ground: A Survey on Dense Mapping and Traversability Methods for Agricultural and Forestry Robots," in IEEE Access, vol. 13, pp. 204285-204314, 2025, doi: 10.1109/ACCESS.2025.3638877.
  - Web page: https://esfandiyar.com/publications/mapping-the-unseen-ground/ · BibTeX: https://esfandiyar.com/publications/mapping-the-unseen-ground/cite.bib · RIS: https://esfandiyar.com/publications/mapping-the-unseen-ground/cite.ris
  - Summary: Scope: dense (not sparse/SLAM-only) 3D mapping for ground robots that must plan motion through unstructured, vegetation-rich farmland and forest. Sensors compared: monocular and stereo cameras, RGB-D / structured light, mmWave radar and LiDAR, by cost, accuracy, resolution, range, compute and robustness to the environment. Map representations reviewed: grids and voxels, octrees (OctoMap), surfels, meshes, TSDF/ESDF (Voxblox, Nvblox), NDT-OM and neural implicit maps, classified by data structure, cell content and hierarchy. Agricultural and forestry adaptations: volumetric and multi-resolution mapping, semantic and collaborative (multi-robot, air–ground) mapping, depth completion and sensor fusion under foliage. Solid ground under vegetation: statistical terrain models, proprioceptive–exteroceptive fusion, radar that penetrates foliage, and self-supervised learning from the robot's own trajectory, feeding traversability maps. Main trade-off identified: memory efficiency vs. map fidelity vs. real-time performance; future work points to uncertainty-aware and active perception and scalable cloud-based multi-robot mapping.
- [GNSS-Based Driver Assistance for Charging Electric City Buses: Implementation and Lessons Learned from Field Testing](https://esfandiyar.com/publications/gnss-driver-assistance-electric-buses/paper.md): Iman Esfandiyar et al., Remote Sensing 15(11), 2938, 2023. doi:10.3390/rs15112938. Preferred citation: Esfandiyar, I.; Ćwian, K.; Nowicki, M.R.; Skrzypczyński, P. GNSS-Based Driver Assistance for Charging Electric City Buses: Implementation and Lessons Learned from Field Testing. Remote Sens. 2023, 15, 2938. https://doi.org/10.3390/rs15112938
  - Web page: https://esfandiyar.com/publications/gnss-driver-assistance-electric-buses/ · BibTeX: https://esfandiyar.com/publications/gnss-driver-assistance-electric-buses/cite.bib · RIS: https://esfandiyar.com/publications/gnss-driver-assistance-electric-buses/cite.ris
  - Summary: Problem: drivers of long electric city buses struggle to dock precisely under pantograph chargers at bus stops and depots, often needing several attempts. Approach: a retrofittable ADAS using two GNSS receivers with RTK corrections delivered over the mobile (GSM/LTE) network for position and heading, an optimisation-based path planner inside obstacle-free bounding rectangles, feedback control, and a driver HMI, built on ROS. Field deployment: evaluated on a Solaris Urbino 12 Electric bus with the public transport operator in Poznań, Poland, reaching Technology Readiness Level 9 (proven in operational environment). Finding: dual-receiver RTK-GNSS was precise enough for docking; failed manoeuvres in all recorded cases were caused by drivers not following the ADAS guidance. Limitation: accuracy degrades near high-rise buildings and dense tree canopy (loss of RTK FIXED), while correction delivery over GSM/LTE stayed reliable regardless of environment and weather. Future work: HMI integrated into the dashboard, and affordable LiDAR for obstacle monitoring and re-planning.

## Projects

- [Modular Robotics Software Platform for Autonomous Agricultural Machinery](https://esfandiyar.com/projects/modular-robotics-platform-agricultural-machinery/) (2026 - now, Technical manager): A modular software stack for autonomous farm machines: perception, navigation and fleet tooling.
- [ICOS: Towards a Functional Continuum Operating System](https://esfandiyar.com/projects/icos-functional-continuum-operating-system/) (2022 - 2025, Technical manager, autonomous farm-robot software developer): ICOS is an EU Horizon Europe project that designed a meta operating system (metaOS) spanning IoT, edge and cloud. At Łukasiewicz PIT, Iman built the project's autonomous farm-robot use case and resilient edge-to-cloud communication for agricultural robots.
- [Agriverse: Synthetic-Data Vision for Crop & Weed Detection](https://esfandiyar.com/projects/agriverse-synthetic-data-crop-weed-detection/) (2025, Project owner, technical manager): A computer-vision pipeline for crop and weed detection, trained on synthetic data generated from digital twins of fields.
- [AgrifoodTEF: Test & Experiment Facilities for Agri-Food AI](https://esfandiyar.com/projects/agrifoodtef-test-experiment-facilities-agri-food-ai/) (2022 - 2025, Chief specialist, plant detection and autonomous farm-robot development): AgrifoodTEF is an EU Digital Europe test and experiment facility for validating AI and robotics solutions in the agri-food domain. Iman worked on plant detection and autonomous farm-robot development.
- [ADAS for Articulated City Buses](https://esfandiyar.com/projects/adas-articulated-city-buses/) (2019 - 2021, Research assistant, system developer): An advanced driver-assistance system (ADAS) for precise manoeuvres of non-articulated and articulated Solaris city buses, including docking under pantograph chargers: GNSS-based guidance, state estimation and control, validated in field tests.
- [Autonomous Drone Swarms](https://esfandiyar.com/projects/autonomous-drone-swarms/) (2021 - 2024, Team lead, product manager): Reconfigurable drone-swarm technology for photovoltaic-farm inspection, drone shows and digital security of critical infrastructure. Iman led the development of the swarm management and monitoring software.

## Optional

- [Full text of all papers in one file](https://esfandiyar.com/llms-full.txt)
- [Curriculum vitae (PDF)](https://esfandiyar.com/assets/iman_esfandiyar_cv.pdf)
- [Publications page](https://esfandiyar.com/publications/)