Explore CTU's Digital Library

Access research publications, theses and academic works

Faculty of Civil Engineering Faculty of Mechanical Engineering Faculty of Electrical Engineering Faculty of Nuclear Sciences and Physical Engineering Faculty of Architecture Faculty of Transportation Sciences Faculty of Biomedical Engineering Faculty of Information Technology Klokner Institute Masaryk Institute of Advanced Studies Central Library Acta Polytechnica Acta Polytechnica CTU Proceedings CTU Publishing Activity

Recent Submissions

3D Printing Filament Dryer using Peltier Heat Pump
Sušička 3D tiskových vláken s využitím Peltierova článku
(Czech Technical University in Prague, 2026-05-22) Bartoň, Vojtěch; Pučejdl, Krištof; Petrucha, Vojtěch
This work investigates possible improvements to consumer Fused Filament Fabrication (FFF) filament dryers using a Thermoelectric Cooler (TEC), also known as a Peltier heat pump. The main focus is on reducing humidity inside the drying enclosure using silica gel and a TEC-assisted moisture-removal system. The measurements showed that silica gel significantly improves the drying process for common materials such as Polylactic Acid (PLA) and Polyethylene Terephthalate Glycol (PETG). Placing the silica gel on the cooled side of the TEC provided an additional improvement. However, it remains questionable whether this benefit justifies the increased complexity of the system.
Design and Validation of a Sensor Network for Modular Structures Made of Digital Materials
Návrh a ověření senzorové sítě pro modulární struktury z digitálního materiálu
(Czech Technical University in Prague, 2026-05-22) Varšányi, Michal; Zemánek, Jiří; Bortel, Radoslav
Large-scale reconfigurable structures built from digital materials require real-time, multi-node sensor aggregation to enable advanced coordination and control. Existing solutions, however, lack the scalability and determinism demanded by voxel-based dynamical systems. This thesis describes the development of VoxLink, a two-generation custom hardware, firmware, and software stack designed to address this gap. Deployed on FPGAs, VoxLink implements a custom fly-by routing protocol offering deterministic, low-latency operation for multi-node sensor aggregation. The hardware achieves 21 Mbit/s data throughput into the FPGA, with a configurable polling interval. Signal integrity was validated across the final generation, with Q-factors as high as 23.3. A comparative study against sensors deployed in existing voxel structures confirmed close agreement in measurements, with the hardware link operating at over 52 times the bandwidth of a traditional fast mode I2C implementation. The design supports expansion to LVDS signaling and USB 2.0 HS interfacing with a host PC, providing greater noise immunity and higher throughput. These properties position VoxLink as a scalable, high-speed sensing backbone for next-generation voxel-based robotic structures.
Multi-Agent Path Finding with Guaranteed Escape Routes
Multiagentní plánování s garancí únikových cest
(Czech Technical University in Prague, 2026-05-22) Konvalinová, Sofie; Zahrádka, David; Rybecký, Tomáš
In this thesis, we introduce and formalize the Multi-Agent Path Finding (MAPF) problem in the presence of a human, focus- ing on guaranteeing escape routes within shared industrial workspaces operated by autonomous robotic fleets. As the de- ployment of autonomous fleets accelerates, modern MAPF solvers are required to produce movement plans that take into account the safety of humans inside the shared workspace. The implementation employs the Large Neighborhood Search (LNS) algorithm, which facilitated the de- sign of a novel safety validation mecha- nism, hereafter referred to as the Human- Safety MAPF algorithm. At every dis- crete time step, this method verifies that the human maintains at least one unob- structed path to a safety zone. If a con- tinuous path does not exist, the spatial configuration of the fleet is classified as a safety failure. This invokes an iterative re- planningoftherobots, whichrepeatsuntil a valid evacuation route is restored. The experimental evaluation focused on the continuous verification of escape routes and the analysis of the overall solution cost. Subsequently, the results were com- pared against a baseline algorithm lacking these safety guarantees in order to accu- rately quantify the computational penalty introduced by the proposed solution.
Shortest Path Problems in Graphs with Limited Resources
Problémy nejkratší cesty v grafech s omezenými zdroji
(Czech Technical University in Prague, 2026-05-22) Weber, Richard; Průša, Daniel; Pěnička, Robert
This thesis addresses shortest path problems in graphs with resource constraints such as limited fuel or energy. Specifically, we define and examine a new problem, the \textit{Resource-refilling constrained shortest walk} (RR-CSW) problem. This problem builds on existing constrained shortest path problems by allowing the resource constraint to be refilled, and therefore expanding its possibilities of application to many real life problems, such as automated transportation. In this thesis. several algorithmic approaches for solving RR-CSW are proposed. We identify selected subclasses of the problem that admit more efficient solutions, and design corresponding algorithms for both exact and heuristic solving of the general case. A software library implementing these methods is developed, together with an application for generating problem instances, executing solvers, and visualizing results. The system is designed to support experimentation with different algorithmic strategies in a unified environment. Finally, the proposed approaches are evaluated experimentally using randomly generated instances. The benchmarks compare runtime and memory consumption across different solvers and demonstrate the scalability limitations of exact methods as well as the trade-offs of heuristic approaches.
Sensor fusion in network of connected vehicles
Fúze sensorických dat propojených vozidel
(Czech Technical University in Prague, 2026-05-25) Jeřábek, Martin; Švancar, Jan; Báča, Tomáš
This master's thesis focuses on the design, implementation, and evaluation of a distributed sensor fusion framework for formations of autonomous vehicles. The main objective is to maintain reliable state estimation even when one member of the formation experiences a temporary outage of its local sensor measurements. The proposed solution is based on a Distributed Extended Kalman Filter architecture, where each vehicle estimates its own state using a nonlinear single-track vehicle model while simultaneously tracking neighboring vehicles using simpler integrator models. State estimates and relative measurements are shared across the communication network, allowing the vehicles to reach a consensus on the estimated states of individual formation members. The thesis also includes the implementation of a V2X communication pipeline based on the European ITS-G5 standard using on-board units provided by Herman Elektronika. To enable the transmission of larger data structures, a custom message splitting and reconstruction utility was designed for payloads exceeding the standard BTP packet size. The proposed algorithm was first verified in simulation and then integrated into a hardware-in-the-loop testing setup using a real V2X communication channel. The results show that sharing estimates within the network improves estimation accuracy and helps bound the estimation error during sensor outages compared to a purely local estimation approach. The thesis also discusses the limitations of the proposed framework, especially the use of a simplified model for neighboring vehicle tracking and the challenges related to the availability of real-world experimental data.