|
|
Simple Terrain Autonomous Navigation
Novák, Daniel ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
An important ability for light terrain outdoor navigation of an UGV (Un- manned Ground Vehicle) is traversability detection. We focus on creating an evaluation layer for move_base package, which is part of ROS (Robot Operating System), using a lidar sensor. After sorting measured points into grid cells of the costmap, we calculate the traversability of each cell. For this calculation, we have proposed three different functions - the difference between the maximum and minimum point heights, variance of point heights and computation of eigenvalues of the covariance matrix. We tested the pro- posed functions in selected real-world situations with respect of operation usability. As a result, the robot is able to reach a given location in light terrain based on the resulting costmap, preferring paths with lower terrain roughness.
|
|
|
Evolution of robots in a simulated physical environment
Bečvář, Marek ; Mráz, František (advisor) ; Vodrážka, Jindřich (referee)
This work introduces a system for designing and evaluating experiments with evo- lutionary algorithms in 3D-simulated physical environments of the MuJoCo library. Ex- periments allow to develop the control and morphology of robots while using arbitrary user-defined fitness functions. The implementation was designed to be as accessible, un- derstandable, and extendable as possible. The system offers a simple graphical user in- terface allowing a detailed configuration of experiments and a text-based user interface which is convenient for running large amounts of experiments for statistical analysis. The work implements several robots of different complexity, examples of various evolutionary algorithms, and a selection of well-known genetic operators. During experiment design, the architecture of this system allows the combining of implemented operators and tools arbitrarily. This work and the user documentation give simple instructions on how to alter and extend the implementation. 1
|
|
|
Řídicí systém robota pro sběr badmintonových míčků
Červeň, Martin ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
Badminton is a racquet game played on court with shuttles made from feath- ers or plastic. Top players train with many shuttlecocks at once, which are fed by coach from hand. After a short training period, shuttlecocks are scat- tered around the court, which need to be picked up so that coach can feed them from hand. In this thesis we created software for autonomous robot that de- tects shuttlecocks with camera, estimates their position and picks them up. We implemented this as nodes in ROS middleware. During development we created simulated environment in Gazebo, and created plugin that simulates shuttle pick- ing. We also created fully working picking mechanism of real shuttlecocks based on rotary brushes powered by motors, utilising 3D printing. Furthermore, we cre- ated and annotated dataset for object detection of over 2500 images and 18500 objects that we used for training and evaluation of state of the art neural net- work, that detects shuttlecocks from video. As part of our solution we developed ROS nodes that allows us to specify working area and area for filtering detections using RViz interactive markers. 1
|
|
|
Visualisation of planning domains
Vodrážka, Jindřich ; Chrpa, Lukáš (advisor) ; Dokulil, Jiří (referee)
Předložená práce zkoumá možnosti vizualizace plánovacích domén. Náplní této práce je návrh grafické reprezentace pro plánovací domény a následné použití tohoto návrhu v programu, který bude fungovat jako grafický editor. Výsledná grafická reprezentace by měla být snadno převoditelná na klasickou reprezentaci plánovacích domén. Program by měl umožňovat práci s plánovacími doménami v rozsahu typed STRIPS reprezentace. Měl by poskytovat uživateli možnost návrhu plánovací domény a její následný export do jazyka PDDL.
|
|
|
Modelling Planning Problems
Vodrážka, Jindřich ; Barták, Roman (advisor) ; Chrpa, Lukáš (referee)
This thesis deals with the knowledge engineering for Automated Planning. The concept of state variables has been recently used with benefits for representation of planning problems. In this thesis the same concept is used in a novel formalism for planning domain and problem modeling. A proof-of-concept knowledge modeling tool is developed based on the new formalism. This tool is then used for modeling of example classical planning domain to show its capabilities. The export to standard domain modeling language is also implemented in the tool in order to provide connection to existing planning systems.
|
|
|
Graph-based SLAM on Normal Distributions Transform Occupancy Map
Jelínek, Lukáš ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
Recent advances in Normal distributions transform occupancy map (NDT-OM) representation have proven to be a viable option for mapping static as well as dynamic environments. Scan registration methods using NDT maps offer a fast and reliable way of registering two laser scans. In this work, we combine 2D NDT mapping and scan matching with the graph-based representation of simultaneous localization and mapping (SLAM). This novel approach uses NDT mini-maps for partial map storage inside the pose graph nodes. It also includes fast incremental scan matcher for odometry estimation. The scan matcher allows to create larger mini-maps which offer better loop closure validation. This work also presents a novel robust distribution to distribution (D2D)-NDT scan matching. It is used for loop closure registration and validation of correct matches. The implementation can operate as an online algorithm inside the Robot Operating System (ROS) framework. The algorithm was tested on MIT Stata Center datasets. Powered by TCPDF (www.tcpdf.org)
|
|
|
Scene Depth Estimation Based on Odometry and Image Data
Zborovský, Peter ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
In this work, we propose a depth estimation system based on image sequence and odometry information. The key idea is that depth estimation is decoupled from pose estimation. Such approach results in multipurpose system applicable on different robot platforms and for different depth estimation related problems. Our implementation uses various filtration techniques, operates real-time and provides appropriate results. Although the system was aimed at and tested on drone platform, it can be well used on any other type of autonomous vehicle that provides odometry information and video output.
|
|
|
Řídicí systém robota pro sběr badmintonových míčků
Červeň, Martin ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
Badminton is a racquet game played on court with shuttles made from feathers or plastic. Top players train with hundreds of shuttles at once which are fed by coach from hand. After a short training period there are hundreds of shuttles scattered around the court, which need to be arranged in rows so that coach can feed them from hand. In this thesis we created software for autonomous robot that detects shuttlecocks with camera, estimates their position and picks them up. We implemented this as nodes in ROS middleware. During development we created simulated environment in Gazebo simulator where we tested our solution. 1
|
|
|
Search-and-Track Techniques for Autonomous Drone
Patík, Vladimír ; Barták, Roman (advisor) ; Vodrážka, Jindřich (referee)
In this diploma thesis we are proposing solution for the search and track problem an evading ground object by an air pursuer in a 3D environment. In addition to static obstacles, there may be dynamic obstacles in the environment which are not plotted on the map provided to the drone. In order to avoid a collision, a safety method has been proposed which is based on processing the image from the input camera by a neural network. The tracking subtask is solved by a reactive algorithm adapted for movement in a built-up area. The probabilistic search algorithm is based on solving Art Gallery and Orienteering problems using nature-inspired algorithms. All algorithms and procedures are evaluated in a simulated environment on randomly generated maps.
|
|
|
Automatic Point Clouds Merging
Hörner, Jiří ; Obdržálek, David (advisor) ; Vodrážka, Jindřich (referee)
Multi-robot systems are an established research area with a growing number of applications. Efficient coordination in such systems usually requires knowledge of robot positions and the global map. This work presents a novel map-merging algorithm for merging 3D point cloud maps in multi-robot systems, which produces the global map and estimates robot positions. The algorithm is based on feature- matching transformation estimation with a novel descriptor matching scheme and works solely on point cloud maps without any additional auxiliary information. The algorithm can work with different SLAM approaches and sensor types and it is applicable in heterogeneous multi-robot systems. The map-merging algorithm has been evaluated on real-world datasets captured by both aerial and ground-based robots with a variety of stereo rig cameras and active RGB-D cameras. It has been evaluated in both indoor and outdoor environments. The proposed algorithm was implemented as a ROS package and it is currently distributed in the ROS distribution. To the best of my knowledge, it is the first ROS package for map-merging of 3D maps.
|
guest ::
RSS feed.