Skip to content

Kaiserslautern, Germany

Amin Dariani

I work on machine perception for robots that have to operate where the world stops being tidy.

Portrait of Amin Dariani
M.Sc. Automation & Control,
RPTU Kaiserslautern, 2023 – present.

At the moment I am a master’s student in Automation and Control at RPTU Kaiserslautern, writing my thesis on environment perception for off-road autonomous driving using diffusion models. Off-road is the interesting case. There are no lane markings, no map priors worth trusting, and a distribution that shifts with the weather, the season, and the light. It is where perception stops being a solved problem and starts being a question again.

B.Sc. Mechanical Engineering,
Ferdowsi University of Mashhad, 2014 – 2019.

My route here was not a straight line. I trained as a mechanical engineer in Mashhad and spent my undergraduate years in a robotics lab working on motion control for a paraplegic lower-limb exoskeleton — work that became my first publication. Then I built software professionally for four years, front-end and full-stack, in Vancouver and then in Spain — and later, alongside the master’s, at SAP in Walldorf.

I came back to research because the questions I could not put down were all in perception and control. The engineering habits came with me, and they turned out to matter more than I expected: I write research code that other people can actually run.

Previously DFKI, 2024 – 2026.

Before Fraunhofer I spent a year and a half at the German Research Center for Artificial Intelligence (DFKI) on gaze-enabled activity classification — inferring what a person is doing from where they choose to look.

Fig. 1 — 6R articulated arm; forward kinematics by nested joint frames. The accent mark is the end effector.

ICurrently

  • Master’s thesis — environment perception for off-road autonomous driving using diffusion models. RPTU Kaiserslautern, ongoing.
  • Working Student Researcher — explainable AI. Fraunhofer IOSB, Karlsruhe.
  • Looking ahead — I am looking for a doctoral position in machine perception, learning for robotics, or human–robot interaction, starting after my thesis.

IIThree threads

Perception under uncertainty

Generative models are unusually good at representing what a scene could plausibly be. My thesis asks whether that helps a vehicle read terrain it has never seen before.

Explanations that hold up

A model that cannot account for itself is hard to trust and harder to debug. My work at Fraunhofer IOSB sits in the gap between an explanation that satisfies a metric and one that satisfies a person.

Robots and the people near them

Exoskeletons, gaze, cooperative control. The systems I keep returning to are the ones where a machine has to read a human’s intent rather than follow a waypoint.

Read about the research →

IIIAway from the desk

I am a slow reader of philosophy and history, a fast and mediocre chess player, and I have spent more hours than I can defend on a badminton court. A few of the other things — the bow, the sabre, the horse, the dance floor — are on a page of their own.

IVGet in touch

I am glad to hear from anyone working on related problems — and especially from groups with doctoral openings.

linkedin.com/in/amin-Dariani