Search

Exploration is a fundamental problem in robotics that requires robots to navigate through unknown environments to autonomously gather information about their surroundings while executing collision-free paths. In this project, we propose a method for producing smooth paths during the exploration process in indoor environments using UAVs to improve battery efficiency and enhance the quality of pose estimation. The developed framework is built by merging two approaches that represent the state of the art in the field of autonomous exploration with UAVs. The overall exploration logic is given by GLocal, a paper that introduces a hybrid, i.e. both sampling-based and frontier-based, framework that is able to cope with the issue of odometry drift when exploring indoor environments due to the absence of absolute localization, e.g. through GNSS. The second approach is FUEL, which introduces a frontier-based exploration methodology which computes the ’drones path as an optimized non-uniform B-Spline. The framework described in this thesis borrows the optimized B-Spline trajectory generation from FUEL and implements it in GLocal. To do this, the original cost function defined by GLocal for each exploration viewpoint was modified and the resulting samples were used to select the initial control points of the B-Spline. Furthermore, we extended the underlying state machine governing the entire algorithm and we revisited the original re-planning logic. The presented system is evaluated in various simulated environments, showcasing the advantages and disadvantages of this method. These evaluations demonstrate its improved state estimation performance and absolute observed volume, albeit at the expense of longer traveled trajectories in big and complex environments., Utforskning är ett grundläggande problem inom robotteknik som kräver att robotar navigerar genom okända miljöer för att autonomt samla in information om sin omgivning samtidigt som de utför kollisionsfria banor. I det här projektet föreslår vi en metod för att producera jämna banor under utforskningsprocessen i inomhusmiljöer med hjälp av UAV:er för att förbättra batterieffektiviteten och förbättra kvaliteten på posestimeringen. Det utvecklade ramverket bygger på en sammanslagning av två metoder som representerar den senaste tekniken inom autonom utforskning med UAV:er. Den övergripande utforskningslogiken ges av GLocal, en artikel som introducerar en hybrid, i.e. både samplingsbaserad och gränsbaserad, ram som kan hantera problemet med odometridrift vid utforskning av inomhusmiljöer på grund av frånvaron av absolut lokalisering, e.g. genom GNSS. Den andra metoden är FUEL, som introducerar en gränsbaserad utforskningsmetod som beräknar drönarens bana som en optimerad icke-uniform B-Spline. Ramverket som beskrivs i denna avhandling lånar den optimerade B-Spline-banegenereringen från FUEL och implementerar den i GLocal. För att göra detta modifierades den ursprungliga kostnadsfunktionen som definierades av GLocal för varje utforskningspunkt och de resulterande samplen användes för att välja de initiala kontrollpunkterna för B-Spline. Dessutom utökade vi den underliggande tillståndsmaskinen som styr hela algoritmen och vi reviderade den ursprungliga logiken för omplanering. Det presenterade systemet utvärderas i olika simulerade miljöer, vilket visar fördelarna och nackdelarna med denna metod. Dessa utvärderingar visar på förbättrad prestanda för tillståndsuppskattning och absolut observerad volym, om än på bekostnad av längre färdvägar i stora och komplexa miljöer.

This project aims to create a robot system that can accurately figure out if the throws are successful. This can help make various industrial tasks more efficient. The system uses implemented methods to process data from fisheye camera data and depth sensor data, to check the quality of the throws. The main goal is to find out if the thrown object reaches its target or not, with more advanced tasks including predicting its path when frames are lost or not tracked properly.To put the system together the Robot Operating System (ROS) was used for handling data and processing, as well as different tools and techniques, like bag files and OpenCV. A variety of methods and algorithms were used to apply background subtraction, clustering, curve fitting, marking objects and drawing the path they take in the air. The depth sensor data processing is included to make up for the limitations of 2D camera data, providing more accurate and reliable tracking of thrown objects.

This project aims to create a robot system that can accurately figure out if the throws are successful. This can help make various industrial tasks more efficient. The system uses implemented methods to process data from fisheye camera data and depth sensor data, to check the quality of the throws. The main goal is to find out if the thrown object reaches its target or not, with more advanced tasks including predicting its path when frames are lost or not tracked properly.To put the system together the Robot Operating System (ROS) was used for handling data and processing, as well as different tools and techniques, like bag files and OpenCV. A variety of methods and algorithms were used to apply background subtraction, clustering, curve fitting, marking objects and drawing the path they take in the air. The depth sensor data processing is included to make up for the limitations of 2D camera data, providing more accurate and reliable tracking of thrown objects.

Exploration is a fundamental problem in robotics that requires robots to navigate through unknown environments to autonomously gather information about their surroundings while executing collision-free paths. In this project, we propose a method for producing smooth paths during the exploration process in indoor environments using UAVs to improve battery efficiency and enhance the quality of pose estimation. The developed framework is built by merging two approaches that represent the state of the art in the field of autonomous exploration with UAVs. The overall exploration logic is given by GLocal, a paper that introduces a hybrid, i.e. both sampling-based and frontier-based, framework that is able to cope with the issue of odometry drift when exploring indoor environments due to the absence of absolute localization, e.g. through GNSS. The second approach is FUEL, which introduces a frontier-based exploration methodology which computes the ’drones path as an optimized non-uniform B-Spline. The framework described in this thesis borrows the optimized B-Spline trajectory generation from FUEL and implements it in GLocal. To do this, the original cost function defined by GLocal for each exploration viewpoint was modified and the resulting samples were used to select the initial control points of the B-Spline. Furthermore, we extended the underlying state machine governing the entire algorithm and we revisited the original re-planning logic. The presented system is evaluated in various simulated environments, showcasing the advantages and disadvantages of this method. These evaluations demonstrate its improved state estimation performance and absolute observed volume, albeit at the expense of longer traveled trajectories in big and complex environments., Utforskning är ett grundläggande problem inom robotteknik som kräver att robotar navigerar genom okända miljöer för att autonomt samla in information om sin omgivning samtidigt som de utför kollisionsfria banor. I det här projektet föreslår vi en metod för att producera jämna banor under utforskningsprocessen i inomhusmiljöer med hjälp av UAV:er för att förbättra batterieffektiviteten och förbättra kvaliteten på posestimeringen. Det utvecklade ramverket bygger på en sammanslagning av två metoder som representerar den senaste tekniken inom autonom utforskning med UAV:er. Den övergripande utforskningslogiken ges av GLocal, en artikel som introducerar en hybrid, i.e. både samplingsbaserad och gränsbaserad, ram som kan hantera problemet med odometridrift vid utforskning av inomhusmiljöer på grund av frånvaron av absolut lokalisering, e.g. genom GNSS. Den andra metoden är FUEL, som introducerar en gränsbaserad utforskningsmetod som beräknar drönarens bana som en optimerad icke-uniform B-Spline. Ramverket som beskrivs i denna avhandling lånar den optimerade B-Spline-banegenereringen från FUEL och implementerar den i GLocal. För att göra detta modifierades den ursprungliga kostnadsfunktionen som definierades av GLocal för varje utforskningspunkt och de resulterande samplen användes för att välja de initiala kontrollpunkterna för B-Spline. Dessutom utökade vi den underliggande tillståndsmaskinen som styr hela algoritmen och vi reviderade den ursprungliga logiken för omplanering. Det presenterade systemet utvärderas i olika simulerade miljöer, vilket visar fördelarna och nackdelarna med denna metod. Dessa utvärderingar visar på förbättrad prestanda för tillståndsuppskattning och absolut observerad volym, om än på bekostnad av längre färdvägar i stora och komplexa miljöer.

Newell's simplified kinematic wave model has been widely used in studying traffic dynamics in a road network. However, it cannot account for First-In-First-Out violation among vehicles on multi-lane roads. In this study, we present an extension of the model in the Lagrangian coordinates. In particular, we define a new variable of a vehicle's order as its corresponding cumulative flows at different times and locations. The new kinematic wave model is consistent at the aggregate level with Newell's simplified kinematic wave model, but allows for FIFO violation among individual vehicles. Based on the new model, we then present two algorithms to estimate vehicles’ trajectories from Eulerian data (cumulative flows provided by loop detectors) and Lagrangian data (the vehicle's entry and exit times provided by reidentification technologies) at two boundaries of a road segment. In the first algorithm, we assume that vehicles follow the First-In-First-Out (FIFO) principle on a multi-lane road, and their orders are constant and equal the average of the entry and exit orders; in the second algorithm, we introduce a linear order-changing model to interpolate a vehicle's orders at different times according to its entry and exit times. With Next Generation Simulation (NGSIM) datasets, we demonstrate that both algorithms are effective, but the second one, with an estimation error of around 10%, greatly outperforms the first one, with an estimation error of around 13%. This verifies the advantage of incorporating FIFO violation in the new model.

Newell's simplified kinematic wave model has been widely used in studying traffic dynamics in a road network. However, it cannot account for First-In-First-Out violation among vehicles on multi-lane roads. In this study, we present an extension of the model in the Lagrangian coordinates. In particular, we define a new variable of a vehicle's order as its corresponding cumulative flows at different times and locations. The new kinematic wave model is consistent at the aggregate level with Newell's simplified kinematic wave model, but allows for FIFO violation among individual vehicles. Based on the new model, we then present two algorithms to estimate vehicles’ trajectories from Eulerian data (cumulative flows provided by loop detectors) and Lagrangian data (the vehicle's entry and exit times provided by reidentification technologies) at two boundaries of a road segment. In the first algorithm, we assume that vehicles follow the First-In-First-Out (FIFO) principle on a multi-lane road, and their orders are constant and equal the average of the entry and exit orders; in the second algorithm, we introduce a linear order-changing model to interpolate a vehicle's orders at different times according to its entry and exit times. With Next Generation Simulation (NGSIM) datasets, we demonstrate that both algorithms are effective, but the second one, with an estimation error of around 10%, greatly outperforms the first one, with an estimation error of around 13%. This verifies the advantage of incorporating FIFO violation in the new model.

Robotic catching of thrown objects is one of the common robotic tasks, which is explored in several works. This task includes subtask of tracking and forecasting the trajectory of the thrown object. Here we propose an algorithm for estimating future trajectory based on video signal from two cameras. Most of existing implementations use deterministic trajectory prediction and several are based on machine learning. We propose a combined forecasting algorithm where the deterministic motion model for each trajectory is generated via the genetic programming algorithm. Genetic programming is implemented on C++ with use of CUDA library and executed in parallel way on the graphical processing unit. Parallel execution allow genetic programming in real time. Numerical experiments with real trajectories of the thrown tennis ball show that the algorithm can forecast the trajectory accurately. © 2016

The objective of the present TFG is to analyze, using different models, how the traffic flow in a freeway is affected by the implementation of VSL zones, ramp metering, HOV lanes or others.

Досліджується вплив на точність траєкторної оцінки різних способів лінеаризації при синтезі алгоритмів одноетапної оптимальної обробки даних вимірювань двопозиційної системи далекомірних радіонавігаційних станцій. Визначаються вирази для статистичних характеристик похибок, а також початкові значення для алгоритмів траєкторних оцінок, що дає можливість забезпечити коректність функціонування алгоритмів. Наводяться результати комп’ютерного моделювання і дається порівняльний аналіз точності при використанні трьох різних підходів до синтезу системи траєкторної оцінки.

The Japanese Hayabusa spacecraft will return to Earth in summer, 2010, carrying samples from asteroid Itokawa. Because the sample return capsule will reenter the atmosphere at night, the capsule and surrounding air will appear as a bright light (i.e., "fireball") during the portion of the trajectory with high aerodynamic heating. Kyushu University, in collaboration with the Japan Aerospace Exploration Agency, is developing a ground-based optical sensor system to observe the reentry and estimate the vehicle's trajectory. This paper describes the design and validation currently underway for the proposed system, in preparation for operations in mid-2010. An Extended Kalman Filter (EKF) is used to estimate the capsule's position and velocity, as well as a scale factor on the atmospheric density. Simulations of the EKF show that the capsule's state at the end of the visible portion of the trajectory (i.e., at approximately 25 km altitude) can be estimated with a 1-s uncertainty of 60 m in position and 8 m/s in velocity., QC 20140903

This paper addresses the issue of estimating underwater vehicle trajectories using gyro-Doppler (body-fixed velocities) and acoustic positioning signals (earth-fixed positions). The approach consists of diffusion-based observers processing a whole trajectory segment at a time, allowing the consideration of important practical problems such as different information update rates, outages, and outliers in a very simple framework. Results of contraction theory are used to prove that the observers are convergent, i.e., stable in the incremental sense. Simulation and experimental results are presented to illustrate the potential of application of the method.

This paper addresses the issue of estimating underwater vehicle trajectories using gyro-Doppler (body-fixed velocities) and acoustic positioning signals (earth-fixed positions). The approach consists of diffusion-based observers processing a whole trajectory segment at a time, allowing the consideration of important practical problems such as different information update rates, outages, and outliers in a very simple framework. Results of contraction theory are used to prove that the observers are convergent, i.e., stable in the incremental sense. Simulation and experimental results are presented to illustrate the potential of application of the method.

This paper addresses the issue of estimating underwater vehicle trajectories using gyro-Doppler (body-fixed velocities) and acoustic positioning signals (earth-fixed positions). The approach consists of diffusion-based observers processing a whole trajectory segment at a time, allowing the consideration of important practical problems such as different information update rates, outages, and outliers in a very simple framework. Results of contraction theory are used to prove that the observers are convergent, i.e., stable in the incremental sense. Simulation and experimental results are presented to illustrate the potential of application of the method.

Tato diplomová práce se zaměřuje na návrh a experimentální ověření systému pro detekci jízdních pruhů, určování trajektorie a polohy vozidla. Cílem bylo vyvinout systém složený z algoritmů pro dílčí části systému. K tomuto účelu bylo využito dat nashromážděných kamerou ZED 2, modelu neuronové sítě U-Net a prvků počítačového vidění ke snížení falešně pozitivních detekcí pomocí časového okna. K určení polohy vozidla vůči trajektorii byly využity trigonometrické výpočty a parametry kamery. Součástí práce bylo také rozšíření datové sady TuSimple o data nashromážděná kamerou. Experimentální ověření prokázalo funkčnost systému s vysokou spolehlivostí detekce v jednoduchých modelových situacích typu jízda po rovném úseku silnice. S náročností modelových situací spolehlivost systému klesá. Navzdory těmto nedostatkům experimenty ukázaly, že systém je schopen detekovat hranice jízdních pruhů a navrhovat optimální trajektorii vozidla. Algoritmy pro určení trajektorie a polohy vozidla závisí na úvodní predikci hranic jízdního pruhu, jsou ale funkční a účinné., This thesis focuses on the design and experimental verification of a system for lane detection, trajectory estimation and vehicle position. The goal was to develop a system composed of algorithms with its respective functions. Data collected with ZED 2 camera, the U-Net neural network model, and computer vision were used to reduce false positive predictions using a temporal window. Trigonometric calculations and camera parameters were used to estimate the vehicle’s position relative to the trajectory. One of the outcomes of this thesis is TuSimple dataset extension with the data captured with ZED 2 camera. Experimental verification demonstrated the system's functionality with high detection reliability in simple model situations, such as driving on a straight road segment. As the complexity of the model situations increased, the system's reliability decreases. Despite these shortcomings, the experiments showed that the system is able to detect lane boundaries and estimate an optimal vehicle trajectory. The algorithms for trajectory and vehicle position determination depend on the initial prediction of the lane boundaries, but they are functional and effective.

Tato diplomová práce se zaměřuje na návrh a experimentální ověření systému pro detekci jízdních pruhů, určování trajektorie a polohy vozidla. Cílem bylo vyvinout systém složený z algoritmů pro dílčí části systému. K tomuto účelu bylo využito dat nashromážděných kamerou ZED 2, modelu neuronové sítě U-Net a prvků počítačového vidění ke snížení falešně pozitivních detekcí pomocí časového okna. K určení polohy vozidla vůči trajektorii byly využity trigonometrické výpočty a parametry kamery. Součástí práce bylo také rozšíření datové sady TuSimple o data nashromážděná kamerou. Experimentální ověření prokázalo funkčnost systému s vysokou spolehlivostí detekce v jednoduchých modelových situacích typu jízda po rovném úseku silnice. S náročností modelových situací spolehlivost systému klesá. Navzdory těmto nedostatkům experimenty ukázaly, že systém je schopen detekovat hranice jízdních pruhů a navrhovat optimální trajektorii vozidla. Algoritmy pro určení trajektorie a polohy vozidla závisí na úvodní predikci hranic jízdního pruhu, jsou ale funkční a účinné., This thesis focuses on the design and experimental verification of a system for lane detection, trajectory estimation and vehicle position. The goal was to develop a system composed of algorithms with its respective functions. Data collected with ZED 2 camera, the U-Net neural network model, and computer vision were used to reduce false positive predictions using a temporal window. Trigonometric calculations and camera parameters were used to estimate the vehicle’s position relative to the trajectory. One of the outcomes of this thesis is TuSimple dataset extension with the data captured with ZED 2 camera. Experimental verification demonstrated the system's functionality with high detection reliability in simple model situations, such as driving on a straight road segment. As the complexity of the model situations increased, the system's reliability decreases. Despite these shortcomings, the experiments showed that the system is able to detect lane boundaries and estimate an optimal vehicle trajectory. The algorithms for trajectory and vehicle position determination depend on the initial prediction of the lane boundaries, but they are functional and effective.