Internship Topics
Prof. Helmut Prendinger
Thank you for your interest in our topics!
Objective:
The Objective of the Internship is a tangible result, such as:
-
Research results that stand a chance of publication in a major venue (journal or conference),
whereby the Intern assumes authorship or co-authorship
- Development or co-development of competitive software, or an intelligent system that
can be demonstrated
Paper writing will be supported by Prendinger Lab.
Topic 1:
Research and Development of Algorithms and Designs
for Unmanned Aircraft Systems ("drone") Traffic Management (UTM)
Description: We are developing a prototype of the entire UTM system to manage drone air traffic,
and conduct related simulation studies.
We are participating in a national UTM project, where we develop a “digital twin” of an area in Japan
(Wakkanai) to simulate and study realistic drone traffic.
Specifically, we have studied scalable algorithms for the pre-flight de-confliction phase among
large numbers of UAVs ("drones"). We are working with several algorithms for Multi Agent
Path Finding (MAPF), such as: Cooperative A*, Enhanced Conflict Based Search (ECBS), etc.
Pre-Flight CDR was tested with the Sendai 2030 Model Case, a realistic projection of
deliveries by drone in 2030.
In busy season, more than 21,000 deliveries per day are expected.
Recently, we started to broaden our simulation studies to risk simulation in the UTM context.
We are also interested in market design for UTM, based on ideas from token economy (Web3).
Research and Development:
- Research part:
- We analyze the different concepts and methods for CDR and study the relevant literature
in MAPF and other related fields.
- Then we model and conceive novel approaches for CDR.
- Development part:
- We implement different algorithms for CDR
- We develop a simulator to perform experimental simulations to evaluate the
performance of algorithms. Java is mostly used.
Key references:
-
Sven Seuken, Paul Friedrich, Ludwig Dierks, Market design for drone traffic management. AAAI, 2022
-
Florence Ho, Ruben Geraldes, Artur Goncalves, Bastien Rigault, Benjamin Sportich, Daisuke Kubo, Marc
Cavazza, Helmut Prendinger.
Decentralized multi-agent path finding for UAV traffic management.
IEEE Transactions on Intelligent Transport Systems, Vol. 23, No. 2, 2022, 997-1008
[DOI: https://doi.org/10.1109/TITS.2020.3019397]
Topic 2:
Deep Learning for Object and Action Recognition from Drone Perspective
Description: We are excited to participate in projects with two National Partners and a university:
-
Advanced Robotics Foundation (ARF): precision landing
-
Central Research Institute of Electric Power Industry (CRIEPI): person detection in bad weather condition
- The University of Tokyo: human action recognition and other Deep Learning applications
ARF: In this project, precious goods have to be carried over the Tokyo Bay area.
Our goal is to develop a Deep Learning based component for precision landing on a dedicated landing pad.
This topic relates to research and development of Deep Learning models precisely landing a drone on a landing pad.
CRIEPI: In this project, we want to develop a river patrol drone system.
Our goal is to develop a Deep Learning based component for detecting persons in bad weather conditions,
such as rain, fog, etc. This topic relates to research and development of Deep Learning models
and practical systems for (near) real-time person detection from the drone perspective.
We have already developed the Person-Action-Locator (PAL) system.
The University of Tokyo: In this project, we study human action recognition from the drone perspective,
and other Deep Learning applications.
We are working with several Deep Learning models, such as:
- Convolutional Neural Networks (CNN)
- Recurrent Neural Networks (RNN)
- Etc.
We are using different kind of input sensors:
- RGB camera on drone
- Multispectral (thermal) camera on drone
Research and Development:
- Research part:
- We train our DL models using different settings of hyper parameters.
- Development part:
- We implement different DL models on NVIDIA Xavier and similar hardware platforms.
We use DL frameworks such as Tensorflow or PyTorch.
Key references:
- Simon Speth, Artur Goncalves, Bastien Rigault, Satoshi Suzuki, Mondher Bouazizi, Yutaka Matsuo, Helmut Prendinger.
Deep Learning with RGB and thermal images onboard a drone for monitoring operations.
Journal of Field Robotics, under revision
-
Mohammadamin Barekatain, Miquel Marti, H.-F. Shih, Samuel Murray, Kotaro Nakayama, Yutaka Matsuo, Helmut Prendinger.
Okutama-Action: An aerial view video dataset for concurrent human action detection.
1st Joint BMTT-PETS Workshop on Tracking and Surveillance, in conj. with CVPR, Honolulu, Hawaii, USA, 2017.7
[DOI: https://doi.org/10.1109/CVPRW.2017.267]
Non-technical overview paper:
- Ruben Geraldes, Artur Goncalves, Tin Lai, Mathias Villerabel, Wenlong Deng, Ana Salta, Kotaro Nakayama,
Yutaka Matsuo, and Helmut Prendinger.
UAV-based situational awareness system using Deep Learning.
IEEE Access, 2019.12, Vol. 7, Issue 1, 122583-122594 (JCR Impact Factor: 4.098)
[DOI: https://doi.org/10.1109/ACCESS.2019.2938249]
Topic 3:
Bitcoin Market Prediction
Description: We aim to understand the predictability of cryptocurrencies, such as Bitcoin.
Our goal is to develop a Deep Learning model for predicting the price of Bitcoin.
We want to consider several Deep Learning models, such as:
- Recurrent Neural Networks (RNN), e.g., GRU, LSTM
We plan to use different kind of inputs:
- Blockchain-based
- Sentiment-based
-
Etc.
Research and Development:
- Research part:
- We train our DL models using different settings of hyper parameters.
- Development part:
- We implement different DL models.
Key references:
- Patrick Jaquart, David Dann, Christof Weinhardt.
Short-term bitcoin market prediction via machine learning.
Journal of Finance and Data Science, Vol. 7, 2021, 45-66
E-mail: helmut @ nii [Address format is: username@nii.ac.jp]
Personal Website: http://research.nii.ac.jp/~prendinger/
LOOKING FORWARD TO HEARING FROM YOU!