PhD Research Fellow in Sensor fusion for perception, collision avoidance and navigation towards autonomous systems
Deadline: 15.09.2020
University of Agder
The University of Agder has more than 1400 employees and 13 000 students. This makes us one of the largest workplaces in Southern Norway. Our staff research, teach and disseminate knowledge from a variety of academic fields. Co-creation of knowledge is our common vision. We offer a broad range of study programmes in many fields. We are situated at two modern campuses in Kristiansand and Grimstad respectively.
We are an open and inclusive university marked by a culture of cooperation. The aim of the university is to further develop education and research at a high international level.
About the position
A fixed-term 100 % position is available at the University of Agder, Faculty of Engineering and Science as a PhD Research Fellow in 3D Sensors and Autonomous systems. The position is affiliated to the Department of Engineering Sciences (Mechatronics group), Senter for forskningsdrevet innovasjon (SFI Offshore Mechatronics) and industrial partners, starting September 2020 and ending September 2023 (with the possibility of one-year extension, with additional teaching assistance duties). This position is located at Campus Grimstad.
Responsibilities
This PhD is strongly tied to academic, research centers and industries where the research fellow, on one hand, will benefit intellectually from strong interaction between academic and well-equipped research centers while gaining exposure to the industries at the same time, thus making a balance between academic and industry. Some of the partners and collaborators with SFI project are – NORCE, NOV, ABB, UiA, NTNU among others. Complete list of partners.
The Mechatronics group is part of Department of Engineering Sciences focusing on 3D sensors and autonomous system, and robotics, focusing on developing various systems, sub-systems and applications related to autonomous vehicle, autonomous navigation in maritime, oil and gas to name a few. This group is actively participating in various projects in strong collaborates with industries and various research centers. Recently it has been home to Top Research Center Mechatronics.
The objective of this PhD thesis is to design, implement and validate sensor fusion and machine learning (ML) algorithms at perception layer for scene perception, object detection and tracking, collision avoidance and navigation towards autonomous systems, especially focused on end to end autonomous vehicle i.e. mapping the sensor data to the control of steering of the vehicle. Among all the other layers, the focus will be on the perception layer to process the signals from different sensors, including Lidar, 3D point-cloud scanners, RGBD cameras, radars or other vision sensors, for scene perception that include the object recognition (exploiting the information received from the vision sensors) and its distance from the vehicle (using the information from Lidar, Radar or other time-of-flight based sensors) in different adverse climatic and environmental conditions. Then, based on all the information, the necessary action is triggered for path planning and control to guide the autonomous system in pre-defined path for collision avoidance. This also includes the interaction of various autonomous vehicles among themselves thus the basic concept of Multi-agent RL (MARL), Multi-objective RL (MORL) will be an added advantage.
To achieve these tasks, various novel data-driven Machine learning (ML) algorithms will be designed and implemented, including processing dynamic 3D point-clouds (e.g. denoising, compression or inference tasks), different architectures of deep neural networks (e.g. CNN, 3D-CNN, GCN/GNN), voxnet-based methods, Inverse reinforcement learning (IRL), deep reinforcement learning (DRL) and its variants with focus on MARL, MORL, and making use of various available libraries (e.g. Yolo, Tensorflow, PyTorch, Keras, OpenCV).
However, this research position is equally open for the candidates who has strong motivation to explore different methodologies to attain autonomy. Depending upon the research interest of the candidates, s/he is free to explore different algorithms, and methodologies.
More specifically the scope of this PhD is:
End-to-end autonomous vehicle.
Use the existing sensor fusion infrastructure (including Lidar, Radar and vision sensors – RGB, Thermal camera) to develop solutions for perception, object recognition, collision avoidance and navigation
Extrinsic and intrinsic calibration of various sensors used in autonomous systems.
Machine Learning algorithms to perform different inferences (information extraction, object recognition, tracking) and autonomous control algorithms for navigation and collision avoidance.
Understanding of fundamental principle, familiar with simulation environment (e.g. OpenAI Gym or similar), and implementing IRL and RL (variants e.g. MORL, MARL)
Strong understanding (both theoretical and practical as basic level) of Reinforcement learning, CNN, Graph Neural Network, Attention Network
A solid academic background with a master’s in electrical engineering, Electronics Engineering, ICT or other related disciplines, is required. It is also possible to apply if the applicant is in the last year of the Master studies and in this case, if the applicant is selected, she or he will start the PhD position once the master’s degree is finished. Recent graduates with industrial experience are also encouraged to apply.
Familiarity with sensors (e.g. Lidar, vision sensors), integration and data acquisition; Robot Operating System (ROS).
Substantial knowledge of most of the following:
Computer vision algorithms, camera calibration, sensor fusion techniques
Analytical and hands-on knowledge on sensors (and sensor fusion)
Machine learning (supervised and non-supervised) algorithms and related frameworks – CNN, RL (multi-agent, multi-objective RL), RNN, GNN, Autoencoders
Optimization techniques, Statistical signal processing
Strong programming skills, mainly in Python, C/C++, MATLAB
Additional programming skills or experience in other relevant tools (e.g. github, bitbucket if any)
Written and spoken English proficiency. Applicants from some countries must document their English proficiency through tests/certificates. Please check this website to see if an English test is required.
The position is remunerated according to the State Salary Scale, salary plan 17.515, code 1017 PhD Research Fellow, NOK 479 600 gross salary per year. A compulsory pension contribution to the Norwegian Public Service Pension Fund is deducted from the pay according to current statutory provisions.
General information
A good working environment is characterised by its diversity. We therefore encourage all qualified candidates to apply for the position, irrespective of gender, age, disability or cultural background. The University of Agder is an IW (Inclusive Workplace).
Women are strongly encouraged to apply for the position.
The successful applicant will have rights and obligations in accordance with the current regulations for the position, and organisational changes and changes in the duties and responsibilities of the position must be expected. Appointment is made by the University of Agder’s Appointments Committee for Teaching and Research Positions.
Short-listed applicants will be invited for interviews. With the applicant’s permission, UiA will also conduct a reference check before appointment. More about the employment process.
In accordance with the Freedom of Information Act § 25 (2), applicants may request that they are not identified in the open list of applicants. The University, however, reserves the right to publish the names of applicants. Applicants will be advised of the University’s intention to exercise this right.
Application
The application and any necessary information about education and experience (including diplomas and certificates) are to be sent electronically. Use the link "Apply for this job".
The following documentation must be uploaded electronically:
Certificates and Diplomas
Updated CV
Project plan (preliminary, not very descriptive). This will be taken into account while evaluating the application
List of publications (if any or link to Google Scholar)
Additional programming skills or experience in other relevant tools (e.g. github, bitbucket if any)
Summary or abstract (approximately 1-2 pages) of the Master Thesis
The applicant is fully responsible for submitting complete digital documentation before the closing date. All documentation must be available in a Scandinavian language or English.
