Supervisor: Linus Nwankwo, M.Sc.;
Univ.-Prof. Dr Elmar Rückert
Start date: 30th October 2023

Theoretical difficulty: mid
Practical difficulty: High

Abstract

This thesis explores machine learning techniques for analysing onboard recordings from the TU Graz Racing Team, a prominent Formula Student team. The main goal is to design and train an end-to-end machine learning model to autonomously discern race courses based on sensor observations.

Further, this thesis seeks to address the following research questions:

• Can track markers (cones) be reliably detected and segmented from onboard recordings?
• Does the delineated racing track provide an adequate level of accuracy to support autonomous driving, minimizing the risk of accidents?
• How well does a neural network trained on simulated data adapt to real-world situations?
• Can the neural network ensure real-time processing in high-speed scenarios surpassing 100 km/h?

Tentative Work Plan

To achieve the objectives, the following concrete tasks will be focused on:

• Thesis initialisation and literature review:
  • Define the scope and boundaries of your work.
  • Study the existing project in [1]  and [2] to identify gaps and methodologies.
  •  
• Setup and familiarize with the simulation environment
  • Build the car model (URDF) for the simulation (optional if you wish to use the existing one)
  • Setup the ROS framework for the simulation (Gazebo, Rviz)
  • Recommended programming tools: C++, Python, Matlab
  •  
• Data acquisition and preprocessing (3D Lidar and RGB-D data)
  • Collect onboard recordings and sensor data from the TU Graz Racing track.
  • Augment the data with additional simulated recordings using ROS, if necessary.
  • Preprocess and label the data for machine learning (ML). This includes segmenting tracks, markers, and other relevant features.
  •  
• Intermediate presentation:
  • Present the results of the literature study or what has been done so far
  • Detailed planning of the next steps
  •  
• ML Model Development:
  •  Design the initial neural network architecture.
  • Train the model using the preprocessed data.
  • Evaluate model performance using metrics like accuracy, precision, recall, etc.
  • Iteratively refine the model based on the evaluation results.
  •  
• Real-world Testing (If Possible):
  • Implement the trained model on a real vehicle’s onboard computer.
  • Test the vehicle in a controlled environment, ensuring safety measures are in place.
  • Analyze and compare the model’s performance in real-world scenarios versus simulations.
  •  
• Optimization for Speed and Efficiency (Optional):
  • Validate the model to identify bottlenecks.
  • Optimize the neural network for real-time performance, especially for high-speed scenarios
  •  
• Documentation and B.Sc. thesis writing:
  • Document the entire process, methodologies, and tools used.
  • Analyze and interpret the results.
  • Draft the thesis, ensuring that at least two of the research questions are addressed.
  •  
• Research paper writing (optional)
  •  

Related Work

[1]   Autonomous Racing Graz, “Enhanced localisation for autonomous racing with high-resolution lidar“, Article by Tom Grey, Visited 30.10.2023.

[2]   Autonomous RC car racing ETH Zürich, “The ORCA (Optimal RC Racing) Project“, Article by Alex Liniger, Visited 30.10.2023.

[3]   P. Cai, H. Wang, H. Huang, Y. Liu and M. Liu, “Vision-Based Autonomous Car Racing Using Deep Imitative Reinforcement Learning,” in IEEE Robotics and Automation Letters, vol. 6, no. 4, pp. 7262-7269, Oct. 2021, doi: 10.1109/LRA.2021.3097345.

[4]   Z. Lu, C. Zhang, H. Zhang, Z. Wang, C. Huang and Y. Ji, “Deep Reinforcement Learning Based Autonomous Racing Car Control With Priori Knowledge,” 2021 China Automation Congress (CAC), Beijing, China, 2021, pp. 2241-2246, doi: 10.1109/CAC53003.2021.9728289.

[5]   J. Kabzan, L. Hewing, A. Liniger and M. N. Zeilinger, “Learning-Based Model Predictive Control for Autonomous Racing,” in IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 3363-3370, Oct. 2019, doi: 10.1109/LRA.2019.2926677.

Supervisor: Univ.-Prof. Dr Elmar Rückert
Start date: 1st March 2023

Involved Company: voestalpine Böhler Aerospace GmbH & Co KG 

Theoretical difficulty: mid
Practical difficulty: mid

Abstract

Geometric data of a requested forging is important as a source to estimate feasibility and offer realistic pricing. However, every bigger deviation in such calculation regarding technical viability costs involved companies’ possible revenue. 

To mitigate this issue and support the technologists and sales department an autoencoder (unsupervised learning) with an attached regression model was developed (pre-existing). Nevertheless, this system still needs adaptation/improvement to meet the operational requirements. 

This bachelor thesis proposes a way to implement an optimization process for adjusting the layer structure and possible scaling of a given autoencoder system. The autoencoder itself uses 3D surface data in form of a “.stl” to create a point cloud in x, y, and z. A docker image containing the autoencoder then extracts the most significant 3D features and provides an estimation for feasibility and price. The focus lies on creating a wrapper function to test different hyperparameters in an automated way. Strategies like random search, grid search, and Bayesian optimization will be applied. The results of the optimized framework will be challenged with the pre-existing autoencoder model.

Tentative Work Plan

To achieve our objective, the following concrete tasks will be focused on:

• Literature research
• Evaluation of the SOTA / the current model
• Identification of network / hyperparameter optimization options
• Model optimization / improvement
• Evaluation and Testing on new data

Supervisor: Univ.-Prof. Dr Elmar Rückert, Qoncept GmbH
Start date: 1st of August 2022

Theoretical difficulty: mid
Practical difficulty: low

Introduction

As direct observations and permanent measurements during steelmaking processes are not possible, modelling has become a powerful tool. The technique of fundamental-based metallurgical modelling is well-established and demonstrates its capabilities in a wide range of applications in modern steelmaking. Following the general trend, data-driven approaches are increasingly used today in various areas of metallurgical modelling, in addition to these  classical fundamental approaches. Depending on the field of application, fundamental-based and data-driven models both have their own advantages and disadvantages.

The overall goal of the present thesis is to combine both models in order to leverage the strengths of  these two different methods. The first step is to apply several different data-driven models and compare them to the metallurgical model to see how they perform differently. In the second phase, various ways of combining data-driven models with the metallurgical model should be investigated. For example, this could be done via a data-driven optimization of its tuning parameters or by replacing them with data-driven models. Also, adding a data-driven residual term to the metallurgical model could be possible. Based on these findings, the third part of the thesis should focus on online learning and methods of how to avoid an off-drifting of the model. The fourth and last section of the thesis should investigate various ways of detecting errors in new data. While point one and two are the main focus of the thesis, point three and four are considered to be optional.

Tentative Work Plan

The following concrete tasks will be focused on:

• Literature research.
• Training of different data-driven models in Python.
• Performance comparison between data-driven models and the metallurgical model.
• Combination of selected data-driven models and the metallurgical model in Python.
• (Optional) Investigate different ways for online learning and live performance evaluation.
• (Optional) Anomaly detection in new data.
• Thesis writing.