WORK WITH ME
Interested in control, learning and real vehicles? I welcome enquiries from motivated students for Master’s theses and research projects.
Explore neural surrogates for discrete-element simulation: reproduce a baseline, assess accuracy and runtime, and investigate a focused extension. The application, data and thesis scope will be agreed together.
Useful background: Python / PyTorch · numerical methods · machine learning
Enquire / apply by email →Investigate adaptive braking control for an e-scooter using reinforcement learning. Start with modelling and simulation, compare against a model-based controller, and evaluate braking performance and robustness. A specific follow-up project can be defined alongside the existing thesis.
Useful background: Control / reinforcement learning · Python or MATLAB · vehicle dynamics
Enquire / apply by email →Choose a topic and use its email link. Send a short introduction, your degree programme, relevant skills, 3–5 sentences on your motivation and your preferred start date. A CV or GitHub link is welcome. We will discuss availability, supervision and the scope in a first conversation.
Topics supervised or co-supervised by Agnes Poks. Availability and the final scope are agreed individually; the neuro-symbolic overtaking thesis is already ongoing.
Extend minimum-time racing-line optimization to elevation and banking; compare the 3D formulation with the existing planar baseline.
Dr. Agnes Poks, Larisa Clement, Philipp Gratzer, Univ. Prof. Ezio Bartocci
Ask about this topic →Build repeatable 3D track maps from LiDAR and IMU data. Benchmark mapping pipelines and investigate one improvement for high-speed racing.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Fuse downward-facing optical-flow measurements with vehicle sensors to estimate sideslip and improve feedback control near the handling limit.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Investigate batched GPU optimization for nonlinear MPC and evaluate solution quality, runtime and real-time feasibility on a racing platform.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Develop RGB-D obstacle detection and tracking with an explicit latency budget, and evaluate accuracy and timing in autonomous racing.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Design multi-sensor pose estimation that detects sensor failures, contains their effects and recovers reliable localization at high speed.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Estimate changing vehicle parameters online with uncertainty information and physically plausible constraints for adaptive racing control.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Localize against an existing 3D LiDAR map, monitor alignment quality and recover globally when local tracking is lost.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Transfer learned liquid vehicle-dynamics models between operating conditions using physical structure and evaluate data efficiency and prediction quality.
Supervision: Dr. Agnes Poks, Larisa Clement, Monika Farsang and Prof. Ezio Bartocci
Ask about this topic →Combine physical vehicle models with liquid neural networks and assess their usefulness for prediction and model predictive control.
Dr. Agnes Poks, Larisa Clement, Monika Farsang, Univ. Prof. Ezio Bartocci
Ask about this topic →Generate follow, overtake and abort trajectory candidates with diffusion models, then use them as behaviour priors for constrained MPCC.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →Combine learned continuous predicates with a symbolic decision layer for safe overtaking, connected to a low-level racing controller.
Dr. Agnes Poks, Larisa Clement, Univ. Prof. Ezio Bartocci
Ask about this topic →I teach and supervise student work in control engineering, mechanics and vehicle dynamics. My responsibilities include course organization, teaching and exercise sessions, examination preparation, and scientific supervision.
Supervision of bachelor research in mechanics and mechatronics.
325.059 · 2026W · PR · 5 h · 10 ECTS · TISS ↗Teaching support in control engineering; prior course organisation, exercises and examinations.
325.038 · 2026W · VU · 3 h · 4 ECTS · TISS ↗Teaching and exercise support in mechanics.
309.021 · 2026W · UE · 3 h · 3 ECTS · TISS ↗Teaching and exercise support in mechanics.
309.022 · 2026W · UE · 3 h · 3 ECTS · TISS ↗Teaching and guidance on vehicle dynamics problems.
325.057 · 2026S · VO · 3 h · 4 ECTS · TISS ↗Course participation includes teaching, exercises and supervision; course ECTS describe the student curriculum, not my individual teaching workload.
The public TISS course programme also lists Bachelor Thesis Mechanics and Mechatronics (325.059), Mechanics exercises 2 (309.022), Project Work in Vehicle System Dynamics or Biomechanics (325.061), and Vehicle System Dynamics and Control (325.057).
TISS course programme ↗Fundamentals of Automatic Control (2019–2021): organisation, exercises and examination preparation. Digital Control (2022): organisation, exercises and examination preparation.
Supervision snapshot reported in the revised FWF application; counts are not added across CV versions.
2 seminar papers, 1 project, 1 individual laboratory exercise, 2 Bachelor’s theses and 1 Master’s thesis. These overlap with later records.
Main supervisor: Agnes Poks · Co-supervisor: Ezio Bartocci
Main supervisor: Agnes Poks · Co-supervisor: Manfred Plöchl
Main supervisor: Agnes Poks
Main supervisor: Agnes Poks · Co-supervisor: Manfred Plöchl
Benjamin Pommer · Diploma thesis, 2023. Advisor: Martin Kozek; co-advisor: Agnes Poks.