Master of Science / Department of Biological & Agricultural Engineering / Daniel Flippo / Robotic vehicles have the potential to play a key role in the future of agriculture. For this to happen designs that are cost effective, robust, and easy to use will be necessary. Robotic vehicles that can pest scout, monitor crop health, and potentially plant and harvest crops will provide new ways to increase production within agriculture. At this time, the use of robotic vehicles to plant and harvest crops poses many challenges including complexity and power consumption. The incorporation of small robotic vehicles for monitoring and scouting fields has the potential to allow for easier integration of robotic systems into current farming practices as the technology continues to develop. Benefits of using unmanned ground vehicles (UGVs) for crop scouting include higher resolution and real time mapping, measuring, and monitoring of pest location density, crop nutrient levels, and soil moisture levels. The focus of this research is the ability of a UGV to scout pest populations and pest patterns to complement existing scouting technology used on UAVs to capture information about nutrient and water levels. There are many challenges to integrating UGVs in conventionally planted fields of row crops including intra-row and inter-row maneuvering. For intra-row maneuvering; i.e. between two rows of corn, cost effective sensors will be needed to keep the UGV between straight rows, to follow contoured rows, and avoid local objects. Inter-row maneuvering involves navigating from long straight rows to the headlands by moving through the space between two plants in a row. Oftentimes headland rows are perpendicular to the row that the UGV is within and if the crop is corn, the spacing between plants can be as narrow as 5”. A vehicle design that minimizes or eliminates crop damage when inter-row maneuvering occurs will be very beneficial and allow for earlier integration of robotic crop scouting into conventional farming practices. Using three fixed HC-SR04 ultrasonic sensors with LabVIEW programming proved to be a cost effective, simple, solution for intra-row maneuvering of an unmanned ground vehicle through a simulated corn row. Inter-row maneuvering was accomplished by designing a transformable tracked vehicle with the two configurations of the tracks being parallel and linear. The robotic vehicle operates with tracks parallel to each other and skid steering being the method of control for traveling between rows of corn. When the robotic vehicle needs to move through narrow spaces or from one row to the next, two motors rotate the frame of the tracks to a linear configuration where one track follows the other track. In the linear configuration the vehicle has a width of 5 inches which allows it to move between corn plants in high population fields for minimally invasive maneuvers. Fleets of robotic vehicles will be required to perform scouting operations on large fields. Some robotic vehicle operations will require coordination between machines to complete the tasks assigned. Simulation of the path planning for coordination of multiple machines was studied within the context of a non-stationary traveling salesman problem to determine optimal path plans.

Dans le cadre du programme ExoMars, l’ESA va déployer un rover sur Mars dont la mission sera de réaliser des prélèvements d’échantillons par forage souterrain et les analyser à l’aide des instruments scientifiques embarqués. Pour atteindre en toute sécurité les différents points d’intérêt où seront effectués ces prélèvements, le rover devra être capable de parcourir plus de 70 mètres par sol (jour martien) tout en respectant les limitations des communications interplanétaires. Les performances des algorithmes de navigation autonome embarqués impacteront directement la réussite scientifique de cette mission. Le premier objectif de cette thèse est d’améliorer les performances de l’architecture de planification de chemin local itératif proposée par le CNES. Tout d’abord, l’utilisation d’un planificateur incrémental de chemin local ”Fringe Retrieving A∗” permettant de réduire la charge de calcul est proposée. Il est complété par l’introduction de tas binaires dans les structures de gestion de la liste de priorité du planificateur de chemin.Ensuite, les manœuvres de rotation sur place pendant l’exécution des trajectoires sont réduites à l’aide d’un planificateur de chemins non-holonomes. Ce planificateur utilise un ensemble de chemins pré-calculés en tenant compte des capacités de braquage du rover. Le second axe de recherche concerne la planification de chemin global d’un rover d’exploration planétaire. Dans un premier temps, la contrainte de mémoire embarquée est détendue et une étude statistique évalue la pertinence d’un planificateur de chemin de type D∗ lite. Dans un deuxième temps, une nouvelle représentation multi-résolution de la carte de navigation est proposée pour stocker de plus grandes zones explorées par le rover sans augmenter l’utilisation de la mémoire embarquée. Cette représentation est utilisée par la suite par un planificateur de chemin global qui réduit automatiquement la charge de calcul en adaptant le sens de recherche en fonction de la forme et de la distribution des obstacles dans l’espace de navigation. / ESA’s ExoMars mission will deploy a 300kg class rover on Mars, which will serveas a mobile platform for the onboard scientific instruments to reach safely desired locations where subsurface drilling and scientific measurements are scheduled. Due to the limited inter-planetary communication constraints, full autonomous on board navigation capabilities are crucial as the rover has to drive over 70 meters per sol(Martian day) to reach designated scientific sites. The core of the navigation softwareto be deployed on the ExoMars rover uses as baseline the autonomous navigation architecture developed by CNES during the last 20 years. Such algorithms are designed to meet the mission-specific constraints imposed by the available spatial technology such as energy consumption, memory, computation power and time costs.The first objective of this thesis is to improve the performance of the successive localpath planning architecture proposed by CNES. First, the use of an increment allocal path planner, Fringe Retrieving A∗, is proposed to reduce the path planning computation load. This is complemented by the introduction of binary heaps in the management structures of the path planner. In-place-turn maneuvers during trajectory execution are further reduced by using a state lattice path planner which encodes the steering capabilities of the rover.The second research direction concerns global path planning capabilities for roboticplanetary exploration. First the onboard memory constraints are relaxed and a studyevaluating the use of a global D∗ lite path planner is performed. Second, a novel multi-resolution representation of the navigation map which covers larger areas atno memory cost increase is proposed. It is further used by a global path planner which automatically reduces the computational load by selecting its search direction based on obstacle shapes and distribution in the navigation space.

Motion planning is defined as the problem of computing a feasible trajectory for an agent to follow. It is a well-studied problem with applications in fields such as robotics, control theory and artificial intelligence. In the last decade there has been an increased interest in algorithms for motion planning under uncertainty where the agent does not know the state of the environment due to, e.g. motion and sensing uncertainties. One approach is to generate an initial feasible trajectory using for example an algorithm such as RRT* and then improve that initial trajectory using local optimization. This thesis proposes a new modification of the RRT* algorithm that can be used to generate initial paths from which initial trajectories for the local optimization step can be generated. Unlike standard RRT*, the modified RRT* generates multiple paths at the same time, all belonging to different families of solutions (homotopy classes). Algorithms for motion planning under uncertainty that rely on local optimization of trajectories can use trajectories generated from these paths as initial solutions. The modified RRT* is implemented and its performance with respect to computation time and number of paths found is evaluated on simple scenarios. The evaluations show that the modified RRT* successfully computes solutions in multiple homotopy classes. Two methods for motion planning under uncertainty, Trajectory-optimized LQG (T-LQG), and a belief space variant of iterative LQG (iLQG) are implemented and combined with the modified RRT*. The performance with respect to cost function improvement, computation time and success rate when following the optimized trajectories for the two methods are evaluated in a simulation study. The results from the simulation studies show that it is advantageous to generate multiple initial trajectories. Some initial trajectories, due to for example passing through narrow passages or through areas with high uncertainties, can only be slightly improved by trajectory optimization or results in trajectories that are hard to follow or with a high collision risk. If multiple initial trajectories are generated the probability is higher that at least one of them will result in an optimized trajectory that is easy to follow, with lower uncertainty and lower collision risk than the initial trajectory. The results also show that iLQG is much more computationally expensive than T-LQG, but that it is better at computing control policies to follow the optimized trajectories.

No description available.

Thesis describes library for algorithm vizualization. It helps to create user interface for application with algorithms. It’s usage is demonstrated on some pathfinding algorithms.These applications are presented on web page.

This work is focused on advanced methods of mobile robot's path planning. The theoretical part describes selected graphical methods, which are useful for speeding up the process of finding the shortest paths, for example through reduction of explored nodes of the state space. In the practical part was created simulate environment in the Python language and in this environment, selected algorithms was implemented.

The goal of this thesis is to nd types of curves which would allow for the construction of a path that could be traversed by a vehicle. It seems that a minimal constraint for such a path is the continuity of curve's curvature. This leads to a closer look at the three types of curves: Clothoids, which are able to smoothly connect straights with arcs of a constant curvature, interpolation quintic splines, which are C2 smooth in the interpolation nodes and -splines, these belong to the family of quintic polynomial curves too, however, they are characterised by the vector of parameters which modies the shape of the curve. The thesis is accompanied by an application allowing for manual construction of the path composed of spline curves.

This thesis is dealing with path planning for autonomous robot in enviromenment with static obstacles. Thesis includes analysis of different approaches for path planning, description of methods utilizing reinforcement learning and experiments with them. Main outputs of thesis are working algorithms for path planning based on Q-learning, verifying their functionality and mutual comparison.

As already clear from the title, this dissertation deals with autonomous locomotive robot path planning, based on machine learning. Robot path planning task is to find a path from initial to target position without collision with obstacles so that the cost of the path is minimized. Autonomous robot is such a machine which is able to perform tasks completely independently even in environments with dynamic changes. Path planning in dynamic partially known environment is a difficult problem. Autonomous robot ability to adapt its behavior to changes in the environment can be ensured by using machine learning methods. In the field of path planning the mostly used methods of machine learning are case based reasoning, neural networks, reinforcement learning, swarm intelligence and genetic algorithms. The first part of this thesis introduces the current state of research in the field of path planning. Overview of methods is focused on basic omnidirectional robots and robots with differential constraints. In the thesis, several methods of path planning for omnidirectional robot and robot with differential constraints are proposed. These methods are mainly based on case-based reasoning and genetic algorithms. All proposed methods were implemented in simulation applications. Results of experiments carried out in these applications are part of this work. For each experiment, the results are analyzed. The experiments show that the proposed methods are able to compete with commonly used methods, because they perform better in most cases.

This work describes path finding among obstacles in 3D space using probabilistic algorithms. Users can create scene in application GUI - define start object, obstacles, goal position and run probabilistic algorithm. The finding path is visualized. The work describes probabilistic algorithm, collision detection and the basics of 3D graphics and shows design and implementation of an application created.