prof. Ing. Jan Faigl, Ph.D.

The topic focuses on autonomous robotic data collection for long-term environmental monitoring and modeling of dynamic spatio-temporal phenomena, such as seismic activity, water systems, forest ecosystems, or environmental changes. The research addresses autonomous sensing using mobile robots (ground or aerial), and distributed sensor networks to efficiently collect informative data for building accurate predictive models. The main research challenges include informative path planning, active sensing, adaptive sampling, persistent monitoring, and uncertainty-aware exploration under motion, energy, and communication constraints. Particular emphasis is placed on multi-robot coordination, distributed data collection, and communication-aware planning for cooperative monitoring in large-scale environments. The work combines methods from robotics, machine learning, optimization, and autonomous systems, with potential validation in realistic outdoor or field deployments.

The topic is dedicated to autonomous data collection tasks motivated by environmental monitoring missions, such as seismic activity modeling or studying ocean or forest phenomena, to build spatio-temporal models of the studied phenomena. The data collection is considered as the autonomous robotic system (robotic vehicle or sensory network) and the problem is to determine the most suitable sensing locations and a cost efficient path for a mobile robot to periodically visit the locations to create a sufficiently precise model of the observed phenomena. http://comrob.fel.cvut.cz/jf

The PhD topic is to investigate how autonomous robots can make their navigation decisions understandable to people—operators, teammates, or bystanders—while they move through dynamic environments. The student will study and design methods that (i) extract the reasons behind path choices from planners and learning-based policies, (ii) communicate those reasons in human-friendly forms (contrastive, intent, “what-if”), and (iii) measure whether explanations improve human mental models, trust, and oversight without degrading safety or performance. The work will combine algorithmic contributions, such as causal/contrastive explanation for A*/POMDP and RL policies, real-time visualization of intent and salient scene factors, and user studies that quantify usefulness, fidelity, and downstream task performance in realistic navigation scenarios.

The topic focuses on learning-based modeling of robotic system dynamics for model-based feedback control, adaptive locomotion, and autonomous motion generation in complex environments. The primary objective is to develop methods that enable robots to learn and exploit dynamic interactions between their bodies and the environment to improve stability, efficiency, and adaptability of motion. Particular emphasis is placed on locomotion generated through robot shape changes and physical interaction with the environment, including crawling, walking, compliant, and soft-body robotic systems. The research addresses learning-based system identification, adaptive control, neural and bio-inspired control architectures, terrain-aware locomotion, and integration of learned dynamics into motion planning and feedback control. The topic may also include reinforcement learning, central pattern generators, sim-to-real transfer, and distributed control for multi-limbed or modular robotic systems.

The topic is focused on learning dynamics of robotic systems towards model-based feedback control. In particular, the main challenges are related to developing learning techniques to locomote by robot shape changes and interactions with the environments, such as crawling or soft-bodies robots.

The topic combines challenges of combinatorial optimization with robotic motion planning in instances motivated by robotic data collection and task planning that requires a solution of multivariable continuous optimization and the solution of the routing problems. The focus of the topic is on a generalization of the existing problem formulations towards solution stability analysis, dynamic and multi-vehicle problems

The topic addresses robotic routing problems arising in autonomous data collection, persistent monitoring, exploration, and cooperative task planning in single- and multi-robot systems. It combines challenges from combinatorial optimization, robotic motion planning, and continuous optimization to develop efficient routing and coordination strategies for autonomous robotic systems operating in dynamic and constrained environments. The research focuses on extending classical routing formulations, such as the Traveling Salesman Problem (TSP), Orienteering Problem (OP), and Vehicle Routing Problem (VRP), toward realistic robotic scenarios involving motion and sensing constraints, uncertainty, communication-limited operation, and distributed multi-vehicle coordination. Particular emphasis is placed on communication-constrained routing, decentralized planning, adaptive replanning, solution stability analysis, and scalable methods for cooperative long-term autonomous missions in changing environments. The topic also considers integrated routing and communication planning, multi-robot task allocation, informative path planning, and coordination of heterogeneous robotic teams under limited bandwidth or intermittent connectivity conditions.

The research aims to develop a robust, passive-based, robot and autonomous system localization by leveraging segmentation-guided visual-language alignment. The approach integrates semantic image segmentation with multimodal VLA models to enhance spatial awareness and scene understanding in unstructured environments. The goal is to extend robustness under challenging illumination by incorporating multispectral sensing, such as near-infrared imaging, to enable reliable localization in low-lighting or visually degraded conditions. By combining pixel-level segmentation with textual or symbolic scene descriptions, the method enables fine-grained, interpretable localization that bridges vision and language domains. The study will explore how learned VLA representations can improve data association, map matching, and relocalization under dynamic and visually ambiguous conditions, contributing to advancements in explainable and human-interpretable robot perception.

The topic focuses on processing and fusing heterogeneous sensor data to create reliable models of the operational environment for autonomous robots and intelligent agents in navigation and exploration missions. The main objective is to develop robust methods for spatial awareness, localization, mapping, and environment understanding that improve the reliability and long-term autonomy of robotic systems operating in dynamic or partially known environments. The research addresses multi-sensor fusion techniques combining data from LiDARs, cameras, inertial sensors, GNSS, radar, or distributed sensing systems to support precise localization and environment modeling. Particular emphasis is placed on reasoning about environment representations, landmark selection and management, uncertainty-aware perception, semantic mapping, and the impact of environment structure on localization accuracy and navigation reliability during long-term autonomous missions. The topic may also include collaborative perception, multi-robot mapping, and communication-aware sharing of spatial information in distributed robotic systems.

In this topic, the main challenges are related to process sensor information to create a model of the operational environment of autonomous agents in autonomous navigation missions. The main goal is to develop methods for improving reliability of the autonomous navigation based on reasoning about the environment model, considering landmarks of the environment utilized for localization and their impact to the precision of the localization in long-term autonomous missions. http://comrob.fel.cvut.cz/jf

The PhD research focuses on developing a unified SLAM and state estimation framework for resource-constrained, bio-inspired robots that exhibit hybrid and discontinuous locomotion dynamics, such as small legged, inchworm-like, or friction-anisotropic climbers. The study aims to achieve accurate vision-based navigation and mapping using low-cost sensors, including monocular or stereo cameras, compact LiDARs, IMUs, magnetometers, and tactile or contact sensors, all operating on lightweight embedded platforms. By integrating probabilistic modeling of hybrid motion phases with multi-sensor fusion and manifold-aware mapping on constrained surfaces, the research will enable robust, real-time SLAM and autonomous navigation for miniature robots in complex, cluttered, or visually challenging environments.