EDI — Elektronikas un datorzinātņu institūts
Atpakaļ uz personālu

Oskars Teikmanis

M. Sc.

Pētnieks

Oskars Teikmanis uzsāka savu darbību Elektronikas un datorzinātņu institūtā 2021. gada septembrī un tika ievēlēts par pētnieku 2023. gada jūlijā. Ieguvis B. Sc. grādu inženierzinātnēs Minhenes Tehniskajā universitātē (TUM) 2016. gadā, Oskars turpināja studijas TUM, lai iegūtu M. Sc. grādu mehatronikā un informācijas tehnoloģijās 2019. gadā. Maģistra darbu rakstīja par robotu lokalizāciju Vācijas aviācijas un kosmosa centrā (DLR). Pēc tam ieguva darba pieredzi autonomās braukšanas jomā Minhenē.

Šobrīd viņa pētnieciskās intereses ir diferencējamās fizikas simulatoru izmantošana konrolējamu sistēmu apmācībā un to pielietošana robotu vadībā.

Saistītās publikācijas:

Teikmanis, O., Leja, L., & Freivalds, K. Applying a Differentiable Physics Simulation to Move Objects with Fluid Streams. (2023). International Workshop on Embedded Digital Intelligence (IWoEDI'2023).

A. C. Hernandez et al., "Searching for Objects in Human Living Environments based on Relevant Inferred and Mined Priors," 2021 European Conference on Mobile Robots (ECMR), 2021, pp. 1-7, doi: 10.1109/ECMR50962.2021.9568792.

Projekti

6 rezultāti

Publikācijas

  • 2024

    Learning to Move Objects With Fluid Streams in a Differentiable Simulation(atveras jaunā logā)

    Kārlis Freivalds, Laura Leja, Oskars Teikmanis

    2024 7th Iberian Robotics Conference, ROBOT 2024, IEEE

    Kopsavilkums

    We introduce a method for contactless manipulation of objects in three-dimensional space using controlled fluid streams. To achieve this, we train a neural network controller in a differentiable simulation and evaluate it in a simulated environment consisting of an 8 × 8 grid of vertical emitters. By carrying out various horizontal displacement tasks such as moving objects to specific positions while reacting to external perturbations, we demonstrate that a controller, trained with a limited number of iterations, can generalise to longer episodes and learn the complex dynamics of fluid-solid interactions. Importantly, our approach requires only the observation of the manipulated object's state, paving the way for the development of physical systems that enable contactless manipulation of objects using air streams.

    Indeksēts Scopus / WoS

  • 2024

    Learning Fluid-Directed Rigid Body Control(atveras jaunā logā)

    Kārlis Freivalds, Oskars Teikmanis, Laura Leja, Rodions Saltanovs, Ralfs Āboliņš

    Machine Learning and the Physical Sciences, NeurIPS

    Kopsavilkums

    We introduce a method for contactless manipulation of objects in three-dimensional space using controlled fluid streams. To achieve this, we train a neural network controller in a differentiable simulation and evaluate it in a simulated environment consisting of a grid of vertical emitters. By carrying out various horizontal displacement tasks such as moving objects to specific positions while reacting to external perturbations, we demonstrate that a controller, trained with a limited number of iterations, can generalise to longer episodes and learn the complex dynamics of fluid-solid interactions. We are in the process of building a floor-standing device that embodies the proposed method, and we describe its key components and design decisions towards displacing and manipulating delicate objects on the device's surface.

  • 2024

    Shaping Flames with Differentiable Physics Simulations(atveras jaunā logā)

    Laura Leja, Oskars Teikmanis, Kārlis Freivalds

    "Machine Learning and the Physical Sciences" at NeurIPS 202, NeurIPS

    Kopsavilkums

    We introduce a method for shaping simulated flames into customisable forms, with the goal of overcoming the limitations of current pyrotechnic systems, which are typically restricted to producing basic fire columns with minimal control over the resulting shape. Our approach leverages differentiable physics and combustion simulations. We demonstrate the use of differentiable physics-based training to produce simulated letter-shaped flames, and take initial steps towards implementing this method on real flame projectors by aligning the simulation with a physical device. The ability to control flame shape would significantly expand the possibilities for stage pyrotechnics and creative applications in performance art.

  • 2024

    Learning to Move Objects with Fluid Streams in a Differentiable Simulation(atveras jaunā logā)

    Kārlis Freivalds, Laura Leja, Oskars Teikmanis

    ROBOT 2024 7th Iberian Robotics Conference, IEEE,

    Kopsavilkums

    We introduce a method for contactless manipulation of objects in three-dimensional space using controlled fluid streams. To achieve this, we train a neural network controller in a differentiable simulation and evaluate it in a simulated environment consisting of an $8\times8$ grid of vertical emitters. By carrying out various horizontal displacement tasks such as moving objects to specific positions while reacting to external perturbations, we demonstrate that a controller, trained with a limited number of iterations, can generalise to longer episodes and learn the complex dynamics of fluid-solid interactions. Importantly, our approach requires only the observation of the manipulated object's state, paving the way for the development of physical systems that enable contactless manipulation of objects using air streams.

    Indeksēts Scopus / WoS

  • 2023

    Automated Vehicle Platform with Connected Driving Capabilities

    Oskars Teikmanis, Aleksandrs Levinskis, Andris Ivars Mackus, Artis Rušiņš, Amr Elkenawy, Marta Tropa, Modris Greitans

    2023 IEEE International Conference on Intelligent Transportation Systems (ITSC), IEEE

    Kopsavilkums

    Augmenting automated vehicles to wirelessly detect and respond to external events before they are detectable by onboard sensors is crucial for developing context-aware driving strategies. To this end, we present an automated vehicle platform, designed with connectivity, ease of use and modularity in mind, both in hardware and software. It is based on the Kia Soul EV with a modified version of the Open-Source Car Control (OSCC) drive-by-wire module, uses the open-source Robot Operating System (ROS and ROS 2) in its software architecture, and provides a straightforward solution for transitioning from simulations to real-world tests. We demonstrate the effectiveness of the platform through a synchronised driving test, where sensor data is exchanged wirelessly, and a model-predictive controller is used to actuate the automated vehicle.

    Indeksēts Scopus / WoS