PhD Fellow - AI for robotization in forestry

Key Information

The PhD fellowship aims to teach robots to understand the forest well enough to be able to navigate and move through it in real time. This is to be achieved using machine learning on LiDAR point clouds and camera images for real-time understanding of the forest environment, and by leveraging this understanding for localization and navigation. A robot builds a picture of its surroundings one LiDAR scan at a time. Each individual scan provides limited information, but together they describe the forest in detail. Traditional methods for analysis and inference are developed to work with completed and quality-assured datasets. A robot, however, must be able to perform estimations based on incomplete observations while in motion.

The PhD project aims to bridge this gap by further developing machine learning models for segmentation and interpretation of forestry point clouds into real-time-based, incremental estimation systems. The work shall support navigation, self-localization, and assessment of traversability in forest terrain.

A bit about the framework of the position

The position is a three-year PhD fellowship with the place of work at NIBIO's headquarters in Ås, 30 km south of Oslo. The candidate will be admitted to a PhD program at Norges miljø- og biovitenskapelige universitet (NMBU) and will work closely with the institute's field robots and associated research activities.

To be qualified for the position, the applicant must not have previously held a PhD fellowship position at NIBIO. The applicant must also not have obtained a PhD or previously held a PhD fellowship funded by Norges forskningsråd.

Applicants who, at the time of employment, do not document skills in Norwegian, Swedish, or Danish at a minimum level of A2, will be offered free Norwegian language training through NIBIO. It is expected that the candidate achieves Norwegian language skills of at least this level during the employment period.

Main tasks of the position

• Develop machine learning architectures that interpret 3D point cloud data incrementally for real-time use, and which gradually build up an understanding of the forest environment as new observations become available.
• Explore compact representations of forest structure and investigate how these can support localization, place recognition, and understanding of the surroundings.
• Adapt models to the characteristics of real-time sensor data, including sparse and unevenly sampled data, motion distortion, and partial observations.
• Implement and evaluate the developed systems on embedded platforms and on NIBIO's robot platforms.

Qualification requirements

• Master's degree or equivalent five-year higher education in computer science, machine learning, robotics, or a related field. The degree must be completed before commencement of the position.
• Qualifications that meet the admission requirements for the relevant PhD program at Norges miljø- og biovitenskapelige universitet (NMBU).
• Experience with implementation, adaptation, and further development of deep neural networks and other machine learning models.
• Good programming skills in Python and experience with modern deep learning frameworks, such as PyTorch or TensorFlow.
• Good skills in English, both oral and written.
• Willingness and physical capacity to participate in fieldwork under varying conditions in Norwegian forest areas.

It is an advantage if you have

• Deep learning for 3D data and sensor data, including point clouds, voxel and sparse representations, transformer models, segmentation, and registration. Experience in developing or adapting model architectures, beyond mere application of existing solutions, is particularly desirable.
• Real-time inference and resource-constrained systems, including quantization, pruning, model compression (distillation), CUDA, and embedded accelerators. Experience with C++ or other system-level programming is an advantage.
• Robotics and state estimation, including SLAM (Simultaneous Localization and Mapping), sensor fusion, place recognition, and ROS/ROS2. Experience with unmanned aerial vehicles (UAV), mobile robots, or machinery within agriculture and forestry will be emphasized.
• Practical experience with development, integration, and use of physical sensor systems, as well as planning and conducting field trials.
• Knowledge of forestry, remote sensing, or ecology.
• Knowledge of Norwegian or another Scandinavian language.

Personal qualities

• Ability to work independently with complex and open research problems, from idea development and method design to implementation, testing, and validation in the field.
• Strong analytical skills and the ability to quickly familiarize oneself with new methods, tools, and subject areas.
• Motivated and engaged in contributing to an ambitious and interdisciplinary research environment.
• Good collaboration and communication skills, and the ability to interact effectively with colleagues and external partners with different professional backgrounds.

We offer

• The opportunity to work with an important societal mission with a focus on sustainability and forward-looking projects
• Challenging, exciting, and varied tasks in interdisciplinary teams
• A flexible everyday life with the opportunity to influence your own tasks
• A strong professional environment with good colleagues
• A generous personnel policy with welfare schemes and a focus on the different life stages of employees

The position is placed in the state salary scale as a PhD fellow, position code 1017, annual salary NOK 555,000 - 635,000. Remuneration is assessed based on education and experience.

Membership in Statens Pensjonskasse, which includes a good occupational pension scheme, occupational injury and group life insurance, as well as offers for home loans at favorable interest rates.

More about NIBIO as an employer can be found here: Jobb i NIBIO

How to apply for the position

The application with CV is filled out electronically and sent via the link on this page, "søk stillingen". Diplomas and certificates are brought to any eventual interview, or attached electronically.

NIBIO conducts background checks on relevant candidates.

Video: https://youtu.be/6mZySoUgD2QOther information

We emphasize diversity in the workplace and encourage qualified candidates to apply - regardless of age, gender, disability, gaps in CV, national or ethnic background.

As a state enterprise, we have the opportunity to provide positive special treatment for applicants who have a disability, gaps in CV, or immigrant background. Positive special treatment includes, among other things, that you will more easily be invited to an interview. Read more about this here.

We point out that information about the applicant may be made public, cf. Act on the Right of Access to Documents in Public Administration (offentleglova), § 25. An applicant may request to be exempted from being listed on the public applicant list. If the request is not granted, the applicant will be notified of this.

About the employer:

NIBIO works with agriculture, food, climate, and environment. We deliver research, management support, and knowledge for application in national preparedness, management, industry, and society at large. NIBIO has its headquarters in Ås in Akershus, has approximately 750 employees, and is present in all regions of the country. NIBIO is owned by Landbruks- og matdepartementet.

Avdeling Driftsteknikk og digitalisering conducts industry-oriented, applied research within forestry production chains from the plant to the industry gate. The purpose is to be a partner to the forestry sector, policymakers, and simultaneously be able to compete in the international research market.