Ing. Robert Pěnička, Ph.D.

This paper presents a family of autonomous Unmanned Aerial Vehicles (UAVs) platforms designed for a diverse range of indoor and outdoor applications. The proposed UAV design is highly modular in terms of used actuators, sensor configurations, and even UAV frames. This allows to achieve, with minimal effort, a proper experimental setup for single, as well as, multi-robot scenarios. Presented platforms are intended to facilitate the transition from simulations, and simplified laboratory experiments, into the deployment of aerial robots into uncertain and hard-to-model real-world conditions. We present mechanical designs, electric configurations, and dynamic models of the UAVs, followed by numerous recommendations and technical details required for building such a fully autonomous UAV system for experimental verification of scientific achievements. To show strength and high variability of the proposed system, we present results of tens of completely different real-robot experiments in various environments using distinct actuator and sensory configurations.

In multi-goal path planning, the task is to find a sequence to visit a set of target locations in an environment using the shortest path. Finding the order can be achieved by solving an instance of the Traveling Salesman Problem (TSP). This requires to determine paths and their lengths between the individual targets, which is solved using robotic path planning. In this paper, we propose a randomized planner for multi-goal path planning for non-holonomic mobile robots. Multiple trees are constructed simultaneously from the targets and expanded by collision-free configurations until they touch each other or obstacles. The trees are expanded using Dubins maneuvers. Therefore, the resulting trajectories satisfy the kinematic constraints of the mobile robots and they can also be used by Unmanned Aerial Vehicles flying at a constant altitude. The efficiency of the proposed planning approach is demonstrated in the multi-goal path planning for environments with tens of targets and compared to state-of-the-art approaches.

In this letter, we propose a novel sampling-based planner for multi-goal path planning among obstacles, where the objective is to visit predefined target locations while minimizing the travel costs. The order of visiting the targets is often achieved by solving the Traveling Salesman Problem (TSP) or its variants. TSP requires to define costs between the individual targets, which — in a map with obstacles — requires to compute mutual paths between the targets. These paths, found by path planning, are used both to define the costs (e.g., based on their length or time-to-traverse) and also they define paths that are later used in the final solution. To enable TSP finding a good-quality solution, it is necessary to find these target-to-target paths as short as possible. We propose a sampling-based planner called Space-Filling Forest (SFF*) that solves the part of finding collision-free paths. SFF* uses multiple trees (forest) constructed gradually and simultaneously from the targets and attempts to find connections with other trees to form the paths. Unlike Rapidly-exploring Random Tree (RRT), which uses the nearest-neighbor rule for selecting nodes for expansion, SFF* maintains an explicit list of nodes for expansion. Individual trees are grown in a RRT* manner, i.e., with rewiring the nodes to minimize their cost. Computational results show that SFF* provides shorter target-to-target paths than existing approaches, and consequently, the final TSP solutions also have a lower cost.

A framework for computing feasible and constrained trajectories for a fleet of quad-rotors leveraging on Signal Temporal Logic (STL) specifications for power line inspection tasks is proposed in this paper. The planner allows the formulation of complex missions that avoid obstacles and maintain a safe distance between drones while performing the planned mission. An optimization problem is set to generate optimal strategies that satisfy these specifications and also take vehicle constraints into account. Further, an event-triggered replanner is proposed to reply to unforeseen events and external disturbances. An energy minimization term is also considered to implicitly save quad-rotors battery life while carrying out the mission. Numerical simulations in MATLAB and experimental results show the validity and the effectiveness of the proposed approach, and demonstrate its applicability in real-world scenarios.

We present a multirotor Unmanned Aerial Vehicle (UAV) control and estimation system for supporting replicable research through realistic simulations and real-world experiments. We propose a unique multi-frame localization paradigm for estimating the states of a UAV in various frames of reference using multiple sensors simultaneously. The system enables complex missions in GNSS and GNSS-denied environments, including outdoor-indoor transitions and the execution of redundant estimators for backing up unreliable localization sources. Two feedback control designs are presented: one for precise and aggressive maneuvers, and the other for stable and smooth flight with a noisy state estimate. The proposed control and estimation pipeline are constructed without using the Euler/Tait-Bryan angle representation of orientation in 3D. Instead, we rely on rotation matrices and a novel heading-based convention to represent the one free rotational degree-of-freedom in 3D of a standard multirotor helicopter. We provide an actively maintained and well-documented open-source implementation, including realistic simulation of UAV, sensors, and localization systems. The proposed system is the product of years of applied research on multi-robot systems, aerial swarms, aerial manipulation, motion planning, and remote sensing. All our results have been supported by real-world system deployment that subsequently shaped the system into the form presented here. In addition, the system was utilized during the participation of our team from the Czech Technical University in Prague in the prestigious MBZIRC 2017 and 2020 robotics competitions, and also in the DARPA Subterranean challenge. Each time, our team was able to secure top places among the best competitors from all over the world.

High-dimensional configuration space is usually searched using sampling-based motion planning methods. The well-known issue of sampling-based planners is the narrow passage problem caused by small regions of the configuration space that are difficult to cover by random samples. Practically, the presence of narrow passages decreases the probability of finding a solution, and to cope with it, the number of random samples has to be significantly increased, which also increases the planning time. By dilating the free space, e.g., by scaling-down or thinning the robot (or obstacles), narrow passages become wider, which allows us to compute an approximate solution. Then, the configuration space can be sampled densely around the approximate solution to find the solution of the original problem. However, this process may fail if the final solution is too far from the approximate one. In this paper, we propose a method to find multiple approximate solutions in the configuration space to increase the chance of finding the final solution. The approximate solutions are computed by repeated search of the configuration space while avoiding, if possible, the already discovered solutions. This enables us to search for distinct solutions leading through different parts of the configuration space. The number of approximate solutions is automatically determined based on their similarity. All approximate solutions are then used to guide the sampling of the configuration space. The performance of the proposed approach is verified in scenarios with multiple narrow passages and the benefits of the method are demonstrated by comparing the results with the state-of-the-art planners.

This paper addresses the perception, control, and trajectory planning for an aerial platform to identify and land on a moving car at 15 km/hr. The hexacopter unmanned aerial vehicle (UAV), equipped with onboard sensors and a computer, detects the car using a monocular camera and predicts the car future movement using a nonlinear motion model. While following the car, the UAV lands on its roof, and it attaches itself using magnetic legs. The proposed system is fully autonomous from takeoff to landing. Numerous field tests were conducted throughout the year‐long development and preparations for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017 competition, for which the system was designed. We propose a novel control system in which a model predictive controller is used in real time to generate a reference trajectory for the UAV, which are then tracked by the nonlinear feedback controller. This combination allows to track predictions of the car motion with minimal position error. The evaluation presents three successful autonomous landings during the MBZIRC 2017, where our system achieved the fastest landing among all competing teams.

Motion planning of 3D solid objects leads to a search in a 6D configuration space. Sampling-based planners randomly sample the configuration space and store the collision-free samples into a graph (roadmap) that can be searched by standard graph-search methods. The well-known issue of the sampling-based planners is the narrow passage problem. Narrow passages are small collision-free regions in the configuration space that are, due to their low volume, difficult to cover by the random samples, which prevents the sampling-based planners to find a path leading through the passages. By decreasing the size of the object, the relative volume of the narrow passages is increased, which helps to cover them more densely. This allows the planner to find an approximate solution, i.e., a solution feasible for the smaller object. The approximate solution can be then used to iteratively guide the sampling in the configuration space, while increasing the size of the object, until a solution for the original object is found. In this paper, we propose a modification of the iterative guiding process. To avoid a situation where the part of the guiding path is too close to obstacles of the configuration space, we shift it away from the obstacles. This requires to estimate the surface of the obstacle region, which is achieved by detecting its boundary configurations during the sampling process. Experiments have shown that the proposed modification outperforms the simple guiding using approximate solutions, as well as other related state-of-the-art planners.

Path planning of 3D solid objects leads to search in a six-dimensional configuration space, which can be solved by sampling-based motion planning. The well-known issue of sampling-based planners is the narrow passage problem, which is caused by the presence of small regions of the configuration space that are difficult to cover by random samples. Guided-based planners cope with this issue by increasing the probability of sampling along an estimated solution (a guiding path). In the case of six-dimensional configuration space, the guiding path needs to be computed in the configuration space rather than in the workspace. Fast computation of guiding paths can be achieved by solving a similar, yet simpler problem, e.g., by reducing the size of the robot. This results in an approximate solution (path) that is assumed to be located near the solution of the original problem. The guided sampling along this approximate solution may, however, fail if the approximate solution is too far from the desired solution. In this paper, we cope with this problem by sampling the configuration space along multiple approximate solutions. The approximate solutions are computed using a proposed iterative process: after a path (solution) is found, it forms a region where the subsequent search is inhibited, which boosts the search of new solutions. The performance of the proposed approach is verified in scenarios with multiple narrow passages and compared with the state-of-the-art planners.

This paper addresses the problem of autonomous cooperative localization, grasping and delivering of colored ferrous objects by a team of unmanned aerial vehicles (UAVs). In the proposed scenario, a team of UAVs is required to maximize the reward by collecting colored objects and delivering them to a predefined location. This task consists of several subtasks such as cooperative coverage path planning, object detection and state estimation, UAV self‐localization, precise motion control, trajectory tracking, aerial grasping and dropping, and decentralized team coordination. The failure recovery and synchronization job manager is used to integrate all the presented subtasks together and also to decrease the vulnerability to individual subtask failures in real‐world conditions. The whole system was developed for the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017, where it achieved the highest score and won Challenge No. 3-Treasure Hunt. This paper does not only contain results from the MBZIRC 2017 competition but it also evaluates the system performance in simulations and field tests that were conducted throughout the year‐long development and preparations for the competition.

Data collection missions are one of the many effective use cases of Unmanned Aerial Vehicles (UAVs), where the UAV is required to visit a predefined set of target locations to retrieve data. However, the flight time of a real UAV is time constrained, and therefore only a limited number of target locations can typically be visited within the mission. In this paper, we address the data collection planning problem called the Dubins Orienteering Problem with Neighborhoods (DOPN), which sets out to determine the sequence of visits to the most rewarding subset of target locations, each with an associated reward, within a given travel budget. The objective of the DOPN is thus to maximize the sum of the rewards collected from the visited target locations using a budget constrained path between predefined starting and ending locations. The variant of the Orienteering Problem (OP) addressed here uses curvature-constrained Dubins vehicle model for planning the data collection missions for UAV. Moreover, in the DOPN, it is also assumed that the data, and thus the reward, may be collected from a close neighborhood sensing distance around the target locations, e.g., taking a snapshot by an onboard camera with a wide field of view, or using a sensor with a long range. We propose a novel approach based on the Variable Neighborhood Search (VNS) metaheuristic for the DOPN, in which combinatorial optimization of the sequence for visiting the target locations is simultaneously addressed with continuous optimization for finding Dubins vehicle waypoints inside the neighborhoods of the visited targets. The proposed VNS-based DOPN algorithm is evaluated in numerous benchmark instances, and the results show that it significantly outperforms the existing methods in both solution quality and computational time. The practical deployability of the proposed approach is experimentally verified in a data collection scenario with a real hexarotor UAV.

This paper focuses on motion planning for information gathering by Unmanned Aerial Vehicle~(UAV) solved as Orienteering Problem~(OP). The considered OP stands to find a path over subset of given target locations, each with associated reward, such that the collected reward is maximized within a limited time of flight. To fully utilize the motion range of the UAV, Hermite splines motion primitives are used to generate smooth trajectories. The minimal time of flight estimate for a given Hermite spline is calculated using known motion model of the UAV with limited maximum velocity and acceleration. The proposed Orienteering Problem with Hermite splines is introduced as Hermite Orienteering Problem~(HOP) and its solution is based on Random Variable Neighborhood Search algorithm~(RVNS). The proposed RVNS for HOP combines random combinatorial state space exploration and local continuous optimization for maximizing the collected reward. This approach was compared with state of the art solutions to the OP motivated by UAV applications and showed to be superior as the resulting trajectories reached better final rewards in all testing cases. The proposed method has been also successfully verified on a real UAV in information gathering task.

This paper introduces the Close Enough Orienteering Problem (CEOP) for planning missions with multi-rotor aerial vehicles considering their maximal velocity and acceleration limits. The addressed problem stands to select the most rewarding target locations and sequence to visit them in the given limited travel budget. The reward is collected within a non-zero range from a particular target location that allows saving the travel cost, and thus collect more rewards. Hence, we are searching for the fastest trajectories to collect the most valuable rewards such that the motion constraints are not violated, and the travel budget is satisfied. We leverage on existing trajectory parametrization based on Bézier curves recently deployed in surveillance planning using unsupervised learning, and we propose to employ the learning in a solution of the introduced multi-vehicle CEOP. Feasibility of the proposed approach is supported by empirical evaluation and experimental deployment using multi-rotor vehicles.

Path planning of 3D solid objects has many practical applications. Sampling-based planners like Rapidly-exploring Random Tree (RRT) tackle the problem by the randomized sampling of the configuration space. The well-known issue of the sampling-based planners is the narrow passage problem. Narrow passages are small collision-free regions in the configuration space that are, due to their low volume, difficult to cover by the random samples, which prevents the sampling-based planners from finding a path leading through the passages. To increase the success rate of the planners, the search in the configuration space can be guided using an approximate solution. An approximate solution is found, for example, considering a smaller object (robot). The performance of this planning concept depends on the way of reducing the size of the object and the ability to find an initial solution. In this paper, we propose a novel technique to reduce the geometry of the object by a combination of triangulation and iterative removal of surface triangles. This technique is suitable for both convex and non-convex objects. In many real-world applications, narrow passages can also contain the initial or goal configurations. We thus propose the extension of the goal region to increase the probability of finding the initial approximate solution. Experiments have shown that the proposed modification outperforms the simple guiding using approximate solutions, as well as other related state-of-the-art planners.

This paper concerns a variant of the Orienteering Problem (OP) that arises from multi-goal data collection scenarios where a robot with a limited travel budget is requested to visit given target locations in an environment with obstacles. We call the introduced OP variant the Physical Orienteering Problem (POP). The POP sets out to determine a feasible, collision-free, path that maximizes collected reward from a subset of the target locations and does not exceed the given travel budget. The problem combines motion planning and combinatorial optimization to visit multiple target locations. The proposed solution to the POP is based on the Variable Neighborhood Search (VNS) method combined with the asymptotically optimal sampling-based Probabilistic Roadmap (PRM*) method. The VNS-PRM* uses initial low-dense roadmap that is continuously expanded during the VNS-based POP optimization to shorten paths of the promising solutions, and thus allows maximizing the sum of the collected rewards. The computational results support the feasibility of the proposed approach by a fast determination of high-quality solutions. Moreover, an experimental verification demonstrates the applicability of the proposed VNS-PRM* approach for data collection planning for an unmanned aerial vehicle in an urban-like environment with obstacles.

In this paper, we propose Greedy Randomized Adaptive Search Procedure~(GRASP) with Path Relinking extension for a solution of a novel problem formulation, the Dubins Team Orienteering Problem with Neighborhoods~(DTOPN). The DTOPN is a variant of the Orienteering Problem~(OP). The goal is to maximize collected reward from a close vicinity of given target locations, each with predefined reward, using multiple curvature-constrained vehicles, such as fixed-wing aircraft or VTOL UAVs with constant forward speed, each limited by route length. This makes it a very useful routing problem for scenarios using multiple UAVs for data collection, mapping, surveillance, and reconnaissance. The proposed method is verified on existing benchmark instances and by real experiments with a group of three fully-autonomous hexarotor UAVs that were used to compare the DTOPN with similar problem formulations and show the benefit of the introduced DTOPN.

Sampling-based motion planners are often used to solve motion planning problems for robots with many degrees of freedom. These planners explore the related configuration space by random sampling. The well-known issue of the sampling- based planners is the narrow passage problem. Narrow passages are small collision-free regions in the configuration space that are, due to their volume, difficult to cover by the random samples. The volume of the narrow passages can be artificially increased by reducing the size of the robot, e.g., by scaling-down its geometry, which increases the probability of placing the random samples into the narrow passages. This allows us to find an approximate solution (trajectory) and use it as a guide to find the solution for a larger robot. Guiding along an approximate solution may, however, fail if this solution leads through such parts of the configuration space that are not reachable or traversable by a larger robot. To improve this guiding process, we propose to compute several approximate solutions leading through different parts of the configuration space, and use all of them to guide the search for a larger robot. We introduce the concept of disabled regions that are prohibited from the exploration using the sampling process. The disabled regions are defined using trajectories already found in the space being searched. The proposed method can solve planning problems with narrow passages with higher success rate than other state-of-the-art planners.

In multi-goal path planning, the task is to find a sequence to visit a set of target locations in an environment. The combinatorial part of the problem (finding the sequence) can be solved as an instance of Traveling Salesman Problem, which requires knowledge about collision-free paths (and distances) between the individual targets. Finding the collision-free paths between the targets is essential in this task. Sampling-based planners like Probabilistic Roadmaps (PRM) and Rapidly-exploring Random Tree (RRT) can be used to find these paths. However, PRM can be computationally demanding, as it attempts to connect each node in the roadmap to its neighbors, regardless of their later usage in the solution. Contrary, RRT is a tree-based planner, and one run can only provide paths starting in the root of the tree (a single target). In this paper, we propose a novel planner for multi-goal path planning. Multiple trees (forest) are constructed simultaneously from the targets and expanded by collision-free configurations until they touch each other or obstacles. Each tree, therefore, does not explore the whole configuration space (as in the case of RRT), and its construction is faster than PRM, as it uses lower number of edges. The efficiency of this new planning approach is demonstrated in the multi-goal path planning in 2D environment with tens of targets and with narrow passages.

The herein studied problem is motivated by practical needs of our participation in the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017 in which a team of unmanned aerial vehicles (UAVs) is requested to collect objects in the given area as quickly as possible and score according to the rewards associated with the objects. The mission time is limited, and the most time‐consuming operation is the collection of the objects themselves. Therefore, we address the problem to quickly identify the most valuable objects as surveillance planning with curvature‐constrained trajectories. The problem is formulated as a multivehicle variant of the Dubins traveling salesman problem with neighborhoods (DTSPN). Based on the evaluation of existing approaches to the DTSPN, we propose to use unsupervised learning to find satisfiable solutions with low computational requirements. Moreover, the flexibility of unsupervised learning allows considering trajectory parametrization that better fits the motion constraints of the utilized hexacopters that are not limited by the minimal turning radius as the Dubins vehicle. We propose to use Bézier curves to exploit the maximal vehicle velocity and acceleration limits. Besides, we further generalize the proposed approach to 3D surveillance planning. We report on evaluation results of the developed algorithms and experimental verification of the planned trajectories using the real UAVs utilized in our participation in MBZIRC 2017.

This paper addresses the recently proposed generalization of the Orienteering Problem (OP), referred to as the Set Orienteering Problem (SOP). The OP stands to find a tour over a subset of customers, each with an associated profit, such that the profit collected from the visited customers is maximized and the tour length is within the given limited budget. In the SOP, the customers are grouped in clusters, and the profit associated with each cluster is collected by visiting at least one of the customers in the respective cluster. Similarly to the OP, the SOP limits the tour cost by a given budget constraint, and therefore, only a subset of clusters can usually be served. We propose to employ the Variable Neighborhood Search (VNS) metaheuristic for solving the SOP. In addition, a novel Integer Linear Programming (ILP) formulation of the SOP is proposed to find the optimal solution for small and medium-sized problems. Furthermore, we show other OP variants that can be addressed as the SOP, i.e., the Orienteering Problem with Neighborhoods (OPN) and the Dubins Orienteering Problem (DOP). While the OPN extends the OP by collecting a profit within the neighborhood radius of each customer, the DOP uses airplane-like smooth trajectories to connect individual customers. The presented computational results indicate the feasibility of the proposed VNS algorithm and ILP formulation, by improving the solutions of several existing SOP benchmark instances and requiring significantly lower computational time than the existing approaches.

Autonomous Micro Aerial Vehicles have the potential to assist in real life tasks involving grasping and transportation, but not before solving several difficult research challenges. In this work, we address the design, control, estimation, and planning problems for cooperative localization, grasping, and transportation of objects in challenging outdoor scenarios. We demonstrate an autonomous team of MAVs able to plan safe trajectories for manipulation of ferrous objects, while guaranteeing inter-robot collision avoidance and automatically creating a map of the objects in the environment. Our solution is predominantly distributed, allowing the team to pick and transport ferrous disks to a final destination without collisions. This result is achieved using a new magnetic gripper with a novel feedback approach, enabling the detection of successful grasping. The gripper design and all the components to build a platform are clearly provided as open-source hardware for reuse by the community. Finally, the proposed solution is validated through experimental results where difficulties include inconsistent wind, uneven terrain, and sandy conditions.

In this paper, we address the data collection planning problem for fixed-wing unmanned aircraft vehicle (UAV) with a limited travel budget. We formulate the problem as a variant of the Orienteering Problem (OP) in which the Dubins airplane model is utilized to extend the problem into the three-dimensional space and curvature-constrained vehicles. The proposed Dubins Airplane Orienteering Problem (DA-OP) stands to find the most rewarding data collection trajectory visiting a subset of the given target locations while the trajectory does not exceed the limited travel budget. Contrary to the original OP formulation, the proposed DA-OP combines not only the combinatorial part of determining a subset of the targets to be visited together with determining the sequence to visited them, but it also includes challenges related to continuous optimization in finding the shortest trajectory for Dubins airplane vehicle. The problem is addressed by sampling possible approaching angles to the targets, and a solution is found by the Randomized Variable Neighborhood Search (RVNS) method. A feasibility of the proposed solution is demonstrated by an empirical evaluation on modified benchmarks for the OP instances to the scenarios with varying altitude of the targets.

In this work, we address data collection planning with an Unmanned Aerial Vehicle (UAV) motivated by surveillance missions in which the UAV is requested to take snapshots at the given set of target locations. In particular, we focus on scenarios where UAV can be modeled by the Dubins airplane model in 3D and the travel budget is limited. In these problems, each target location has associated reward value representing an importance of the target, and thus the studied planning problem is to determine the most valuable targets together with the sequence to their visits such that the length of the data collection trajectory fits the travel budget.

In this paper, we address the Orienteering Problem (OP) for curvature constrained vehicle. For a given set of target locations, each with associated reward, the OP stands to find a tour from a prescribed starting location to a given ending location such that it maximizes collected rewards while the tour length is within a given travel budget constraint. The addressed generalization of the Euclidean OP is called the Dubins Orienteering Problem (DOP) in which the reward collecting tour has to satisfy the limited turning radius of the Dubins vehicle. The DOP consists not only of selecting the most valuable targets and determination of the optimal sequence to visit them, but it also involves the determination of the vehicle’s heading angle at each target location. The proposed solution is based on the Variable neighborhood search technique, and its feasibility is supported by an empirical evaluation in existing OP benchmarks. Moreover, an experimental verification in a real practical scenario further demonstrates the necessity of the proposed direct solution of the Dubins Orienteering Problem.

In this paper, we address the Dubins Orienteering Problem with Neighborhoods (DOPN) a novel problem derived from the regular Orienteering Problem (OP). In the OP, one tries to find a maximal reward collecting path through a subset of given target locations, each with associated reward, such that the resulting path length does not exceed the specified travel budget. The Dubins Orienteering Problem (DOP) requires the reward collecting path to satisfy the curvature-constrained model of the Dubins vehicle while reaching precise positions of the target locations. In the newly introduced DOPN, the resulting path also respects the curvature constrained Dubins vehicle as in the DOP; however, the reward can be collected within a close distant neighborhood of the target locations. The studied problem is inspired by data collection scenarios for an Unmanned Aerial Vehicle (UAV), that can be modeled as the Dubins vehicle. Furthermore, the DOPN is a useful problem formulation of data collection scenarios for a UAV with the limited travel budget due to battery discharge and in scenarios where the sensoric data can be collected from a proximity of each target location. The proposed solution of the DOPN is based on the Variable Neighborhood Search method, and the presented computational results in the OP benchmarks support feasibility of the proposed approach.

In this paper, we address a generalization of the Orienteering Problem (OP) for curvature-constrained vehicles and to problems where it is allowed to collect a reward associated to each target location within a specified distance from the target. The addressed problem combines challenges of the combinatorial optimization of the OP (to select the most rewarding targets and find the optimal sequence to visit them) with the continuous optimization related to the determination of the waypoint locations and suitable headings at the waypoints for the considered Dubins vehicle such that the curvature-constrained path does not exceed the given travel budget and the sum of the collected rewards is maximized. The proposed generalization is called the Close Enough Dubins Orienteering Problem (CEDOP) and novel unsupervised learning approach is proposed to address computational requirements of this challenging planning problem. Based on the presented results, the proposed approach is feasible and provides a bit worse solution of CEDOP than the existing combinatorial approach but with significantly lower computational requirements.

We introduce a novel online replanning method for robotic information gathering by Unmanned Aerial Vehicles~(UAVs) called Reactive Dubins Traveling Salesman Problem~(RDTSP). The considered task is the following: a set of target locations are to be visited by the robot. From an initial information gathering plan, obtained as an offline solution of either the Dubins Traveling Salesman Problem~(DTSP) or the Coverage Path Planning~(CPP), the proposed RDTSP ensures robust information gathering in each given target location by replanning over possible missed target locations. Furthermore, a simple decision making is a part of the proposed RDTSP to determine which target locations are marked as missed and also to control the appropriate time instant at which the repair plan is inserted into the initial path. The proposed method for replanning is based on the Variable Neighborhood Search metaheuristic which ensures visiting of all possibly missed target locations by minimizing the length of the repair plan and by utilizing the preplanned offline solution of the particular information gathering task. The novel method is evaluated in a realistic outdoor robotic information gathering experiment with UAV for both the Dubins Traveling Salesman Problem and the Coverage Path Planning scenarios.

In this paper, we address the Orienteering problém (OP) by the unsupervised learning of the self-organizing map (SOM). We propose to solve the OP with a new algorithm based on SOM for the Traveling salesman problem (TSP). Both problems are similar in finding a tour visiting the given locations; however, the OP stands to determine the most valuable tour that maximizes the rewards collected by visiting a subset of the locations while keeping the tour length under the specified travel budget. The proposed stochastic search algorithm is based on unsupervised learning of SOM and it constructs a feasible solution during each learning epoch. The reported results support feasibility of the proposed idea and show the performance is competitive with existing heuristics. Moreover, the key advantage of the proposed SOM-based approach is the ability to address the generalized OP with Neighborhoods, where rewards can be collected by traveling anywhere within the neighborhood of the locations. This problem generalization better fits data collection missions with wireless data transmission and it allows to save unnecessary travel costs to visit the given locations.

This paper introduces a new concept for flexible motion planning and control of industrial robots. Instead of a closed monolithic architecture, an open service-based framework is proposed. The services can be run hardware-independent on a decentralized IT infrastructure (Cloud) allowing a fast reconfiguration of control modules and their multiple usage in different tasks. The services can be run on a dedicated PC or on individual virtual machines inside a single computing cluster. The proposed service-based framework was implemented and tested for the motion planning of a robotic manipulator. Effects on control performance, availability and scalability are investigated and documented.