Leonard Rychly

Household Robotics, Embedded Machine Learning, Computer Vision, Neural Network Compression and Optimization

Division for Trustworthy Autonomous Systems. Worked on deep reinforcement learning for generating robust behavior strategies in uncertain and partially observable environments. Assessed the generalization capabilities of learned strategies. Worked on autopilot systems for unmanned aerial vehicles. Industry project on robust road user classification for highly automated driving based on camera, lidar and radar data.

Developed a prototype to control high-dynamic, non-linear hydraulic systems using model-free deep reinforcement learning and adaptive control.

Processed and visualized large data sets of indoor climate measurements.

Software testing and evaluation of computer vision algorithms.

Thermal Benchmark Calculation

1. Praktisches Semester

Focus on: AI, Reinforcement Learning, Machine Learning, Computer Vision, Robot Dynamics. Master's Thesis: Decoding of 3D reach and grasp movements from non-invasive EEG signals using spiking neural networks on SpiNNaker Neuromorphic Hardware

Numerically modeling a complex fluid structure interaction problem: quantifying the aerodynamic damping induced by the surrounding air on a vibrating structure.

Thesis: Design and evaluate computer vision techniques to measure the fluid level of small Plexiglas tubes under challenging external lightning conditions. Developed in C++ with OpenCV on a BeagleBone Black.