Welcome Students – cool projects are waiting for you!

Students will find open project work at all levels here – please check regularly.

Prospective PhD students please look up the open positons:  https://human-centered.ai/open-positions

Below you find a selection of open student work in random order in top-level description. Details, schedules and deliverables, as well as individual means of communication (e.g. email contacts, Slack, Discord, or any other) will be arranged directly with the respective supervisor. We want to match the ideally suited students (knowledge, expertise, interest) to the most suitable project.

[1.]  Watch the inaugural lecture: https://www.youtube.com/watch?v=odFFJIgDElw&t=1096s

[2.] Read our Research plan: https://doi.org/10.3390/s22083043

Topics S-01:  To monitor Forest Roads and predict their trafficability

Motivation: According to the Austrian Federal Ministry of Agriculture, Forestry, Regions and Water Management, there are about 170,000 km of forest roads in Austria. In comparison, the length of the Austrian highway network is about 2,215 km (as of the end of 2021), as reported by ASFINAG (Autobahnen- und Schnellstraßen-Finanzierungs-Aktiengesellschaft). It is obvious that there are a lot of opportunities that Artificial Intelligence (AI) and Machine Learning (ML) can be used in combination with robots, drones, airborne laser scanning, weather data, environmental data, and sensor data (e.g., soil sensors, pressure sensors, moisture sensors, etc.) to monitor forest roads and predict their trafficability.

Together with our 980 hectares of experimental forest (Lehrforst) and our robot test park in Tulln we have a lot of opportunties for research and developement tasks at all levels (students projects, bachelor theses, master theses, PhD, etc.), if you are interested contact us directly e.g. via andreas.holzinger AT human-centered.ai

Here are some examples:

Drone-based inspection: drones can be equipped with cameras and sensors to inspect the forest road from the air and detect various features of the road, such as cracks, damage, obstacles, or changes on the road surface. Machine learning algorithms can be used to analyze and process this data to identify patterns and trends to make predictions about the road’s passability. Autonomous drone flight is a hard problem and there are a lot of unsolved problems.

Sensor-based monitoring: ground sensors and other IoT devices can also be placed along the forest road to collect data on various parameters such as temperature, humidity, soil stability, traffic density, and other relevant parameters. This data can also be analyzed by machine learning algorithms to predict the trafficability of the road.

Robot-based inspection: robots can drive either autonomously or remotely along the forest road to collect data on the road’s condition. These robots can be equipped with cameras, sensors, and other devices to capture and assess various features of the road. Machine learning techniques can be used to process the data collected by these robots to make predictions about the passability of the road.

Airborne laser scanning: Airborne laser scanning (ALS) can be used to create 3D models of the surrounding area and road layout using lidar technology. These models can provide information about the terrain shape, road surface, obstacles, damage, vegetation, and other factors that help predict the passability of the forest road.

Weather data and environmental data: Weather and environmental data can be used to improve predictions. For example, the amount of rainfall or snowfall can have an impact on the passability of the road. Combining weather data with other data sources and machine learning techniques can improve predictions of road drivability.

Combination of all the above approaches: Combining the above approaches can also be effective in improving predictions. For example, drone inspection can be used to assess the overall condition of the road, while sensor-based data and robotic inspections can be used to identify specific problems or bottlenecks.

References:

Topics S-02:  Predictive Maintainance – “Machine Health”

Predictive Maintenance for Chainsaws: Chainsaws are a critical tool for forestry operations, and a failure can cause significant disruption. Students can explore the development of an AI-driven predictive maintenance model that uses data from sensors and other sources to predict when a chainsaw is likely to fail. The project can involve the collection of data from various types of chainsaws and the application of machine learning algorithms to predict when maintenance is necessary.

Predictive Maintenance for Heavy Equipment: Heavy equipment such as harvesters and skidders are expensive, and a failure can be costly. Students can explore the development of an AI-driven predictive maintenance model that uses data from sensors and other sources to predict when maintenance is necessary. The project can involve the collection of data from various types of heavy equipment and the application of machine learning algorithms to predict when maintenance is necessary.

Predictive Maintenance for Forest Roads: Forest roads are critical for forestry operations, and a failure can cause significant disruption. Students can explore the development of an AI-driven predictive maintenance model that uses data from sensors and other sources to predict when maintenance is necessary. The project can involve the collection of data from various types of forest roads and the application of machine learning algorithms to predict when maintenance is necessary.

Analysis of Failure Data: Failure data from forestry equipment can be used to identify patterns and trends that can help in the development of predictive maintenance models. Students can explore the application of statistical techniques and machine learning algorithms to failure data from forestry equipment to identify patterns and trends that can inform the development of predictive maintenance models.

Sensor Data Analysis: Sensor data can be used to monitor the performance of forestry equipment and identify potential issues before they become serious. Students can explore the development of algorithms to analyze sensor data from forestry equipment and identify potential issues. The project can involve the collection of sensor data from various types of forestry equipment and the development of machine learning algorithms to analyze the data.

Image Recognition for Maintenance: Images of forestry equipment can be used to identify potential issues that require maintenance. Students can explore the development of image recognition algorithms that can identify potential maintenance issues in images of forestry equipment. The project can involve the collection of images of various types of forestry equipment and the development of machine learning algorithms to analyze the images.

Natural Language Processing for Maintenance: Maintenance logs and other written documentation can be used to identify potential issues that require maintenance. Students can explore. The project can involve the collection of maintenance logs and other written documentation and the development of machine learning algorithms to analyze the data.

Environmental Data Analysis: Environmental data such as temperature, humidity, and precipitation can have an impact on the performance of forestry equipment. Students can explore the development of algorithms to analyze environmental data and identify potential issues that require maintenance. The project can involve the collection of environmental data from various forestry operations and the development of machine learning algorithms to analyze the data.

Data Visualization: Data visualization can be used to communicate the results of predictive maintenance models to stakeholders. Students can explore the development of data visualization tools to communicate the results of predictive maintenance models to stakeholders. The project can involve the development of data visualization tools using various data sets and the assessment of the effectiveness and the usability of the tools in communicating the results to stakeholders.

Integration with Maintenance Management Systems: Predictive maintenance models need to be integrated with maintenance management systems to ensure that maintenance is scheduled and performed when necessary. Students can explore the integration of predictive maintenance models with maintenance management systems. The project can involve the development of a prototype system that integrates a predictive maintenance model with a maintenance management system and the assessment of the effectiveness of the system.

References:

[1] Thyago P. Carvalho, Fabrizzio Soares, Roberto Vita, Roberto Francisco, Joao P Basto & Symone Alcala (2019). A systematic literature review of machine learning methods applied to predictive maintenance. Computers & Industrial Engineering, 137, 106024, doi:10.1016/j.cie.2019.106024.

Topics S-03:  Forestry 5.0 on the example of cable yarding in steep terrain

Route planning optimization: Locate open problems and challenges and find methods to mitigate them, so to optimize the route planning taking into account factors such as terrain, log size, and cable length, and optimize the route to minimize travel time and fuel consumption.

Fuel consumption reduction: Evaluate methods that can reduce fuel consumption during cable yarding operations in steep terrain. Find out how machine learning algorithms can be used to optimize machine settings such as throttle, winch speed, and cable tension, and minimize fuel consumption while maintaining productivity.

Simulators for training: A huge topic is design, development, test and evaluate realistic cable yarding simulators for training purposes. The simulator should be able to replicate the experience of operating a cable yarding machine in steep terrain, and provide trainees with a safe and controlled environment in which to practice their skills.

Machine learning for machine health: Research on possibilities on how we can predict maintenance needs for cable yarding machines. This includes to analyze machine data such as engine temperature, hydraulic pressure, and cable tension, and predict when maintenance will be required to prevent breakdowns and extend machine life.

These challenges are just a starting point, and there are many other areas where AI and machine learning can be applied to improve cable yarding in steep terrain.

Topics S-04:  Forestry 5.0 on the example of cyber-physical systems in the forests

Sample use cases:

Forest monitoring: Students can experiment with cyber-physical systems (sensors) that can monitor the forest for various environmental factors such as temperature, humidity, and air quality. These systems can also detect forest fires and help prevent them.

Forest mapping: Students can experiment with robots that can create 3D digital twins of the forest to better understand the landscape and identify areas that need protection.

Open challenges:

  1. Power management: One of the biggest challenges in developing robots or cyber-physical systems for the forest is power management. These systems need to be self-sufficient and able to operate in areas without access to electricity.
  2. Navigation: Navigation can be difficult in the forest due to obstacles such as trees and uneven terrain. Students can work on developing navigation algorithms and systems that can overcome these challenges.
  3. Robustness: The forest environment can be harsh, with extreme weather conditions, water, and mud. Students can work on developing robots or cyber-physical systems that can withstand these conditions and still operate reliably.
  4. Human-robot interaction: As robots and cyber-physical systems become more prevalent in the forest, it is important to consider how they will interact with humans. Students can work on developing interfaces that allow humans to interact with these systems in a natural and intuitive way.