No description available.

Localization is a critical aspect of robots and their industrial applications, with its major impact on navigation and planning. The goal of this thesis is to improve multi-robot localization by utilizing scan matching algorithms to calculate a corrected pose estimate using the robots' shared laser scan data. The current pose estimate relative to the map is used as the initial guess for the scan matching. This corrected pose is fused using several different localization configurations, such as an Extended Kalman Filter in combination with the Adaptive Monte Carlo Localization algorithm. Simulations showed that localization improved by resetting the Monte Carlo particle filter with the pose estimate generated by the collaborative scan matching. Further, in simulated scenarios, the collaborative scan matching implementation improved the accuracy of typical Monte Carlo Localization configurations. Furthermore, when filtering based on the number of reciprocal correspondences between the scan match output and the target scan, one could extract highly accurate pose estimates. When resetting the Monte Carlo Localization algorithm with the pose estimates, the localization algorithm could successfully recover from severe errors in the global positioning. In future work, additional testing needs to be done using these extracted pose estimates in a dedicated map-based multi-robot localization algorithm.

Convolutional neural networks have been successfully used in action recognition but are usually restricted to operate on Euclidean data, such as images. In recent years there has been an increase in research devoted towards finding a generalized model operating on non-Euclidean data (e.g graphs) and manipulation action recognition on graphs is still a very novel subject. In this thesis a novel graph based deep neural network is developed for predicting manipulation actions and reconstructing graphs from a lower space representation. The network is trained on two manipulation action datasets and uses their, respective, previous works on action prediction as a baseline. In addition, a modular perception pipeline is developed that takes RGBD images as input and outputs a scene graph, consisting of objects and their spatial relations, which can then be fed to the network to lead to online action prediction. The network manages to outperform both baselines when training for action prediction and achieves comparable results when trained in an end-to-end manner performing both action prediction and graph reconstruction, simultaneously. Furthermore, to test the scalability of our model, the network is tested with input graphs deriving from our scene graph generator where the subject is performing 7 different demonstrations of the learned action types in a new scene context with novel objects.

I robotikkursen Autonoma Mekatroniska System på Halmstad Högskola finns ett problem som kan uppstå vid bygget av deras hjulförsedda mobila robotar. Studenter har en tendens att bygga roboten så att den körs med maximal effekt från start till inbromsning. Detta medför att vid test av deras robotar så slits motorerna ut fortare vilket ger en onödig kostnad. För att minska på kostnaden är målet med detta projekt att bygga och reglera en sådan robot med hjälp av en banplanerare. Banplaneraren skapar mjuka övergångar vid start och inbromsning av roboten vilket gör att motorerna inte slits ut lika fort. I detta projekt ska vi utveckla en robot som kan användas som ett pedagogiskt undervisningsmaterial. Roboten ska vara en mall i kursen för studenterna så de kan se att det går bygga en robot som kan köras med mjuka övergångar. Roboten ska både kunna följa ett rörligt objekt med ett bestämt avstånd och har möjlighet att rotera. I detta projekt har även en undersökning av en modellerad och en verklig motor gjorts genom systemidentifiering. Resultatet blev en robot som kan följa ett rörligt objekt med ett bestämt avstånd där start och inbromsning görs mjukt. Med denna robot kan studenter se att det är möjligt att bygga en robot som körs med mjuka övergångar, vilket leder till mindre slitage av motorerna. Stegsvar och bodediagram har plottats fram genom systemidentifieringen där skillnader och likheter mellan verklig och modellerad motor kan ses. / In the robotics course Autonomous Mechatronic Systems at Halmstad University there is a problem that may occur during the construction of their wheeled mobile robots. Students have a tendency to build the robot so that it runs with maximum efficiency from start to braking. This means that when testing their robots, the motors wear out faster, which is providing an unnecessary cost. To reduce the cost, the goal of this project is to build and regulate such a robot using a positionplanner. The positionplanner creates smooth transitions at start and braking of the robot so that the motors does not wear out as fast. In this project, we will develop a robot that can be used as an educational teaching materials. The robot will be a template in the course for the students so they can see that it is possible to build a robot that can run with smooth transitions. The robot should both be able to follow a moving object with a fixed distance and be able to rotate. In this project, an investigation of a modeled and actual motor has been made through a systemidentification. The result was a robot that can follow a moving object with a fixed distance where starting and braking is done smoothly. With this robot, students can see that it is possible to build a robot running with smooth transitions, which leads to less wear on the motors. Stepresponses and bodeplots have been plotted through the systemidentification where differences and similarities between real and modeled motor can be seen.

Den obemannade tröskeln är en uppsats på magisternivå efter avslutade studier på HOP1 vid Försvarshögskolan. Syftet med att skriva den var att försöka ge ett bidrag till anskaffningsprocessen av nya autonoma system till den svenska marinen. Detta gjordes genom att använda de tydligast framträdande dimensionerande teorierna ur svensk militärstrategisk doktrin och skapa ett analysverktyg utifrån detta, tillsammans med andra enligt författaren viktiga svenska kulturella principer, nämligen efterlevnaden av krigets lagar och hänsyn till kostnad. Studien är konstruerad som en kvalitativ, utforskande fallstudie med den svenska marinens undervattensförmåga som studerat fall. Resultaten visar på att marinen kan ha mycket stor nytta av dessa system, men att det initialt är viktigt att få ut system tidigt som direkt stödjer befintliga plattformar i deras verksamhet.

The possibility of using multiple autonomous agents for completing various tasks has gathered a lot of interest in the last years. Using multiple agents does however come with the problems of choosing how to allocate the tasks and how the intelligence of the individual agents should be designed.  In this thesis a market inspired auction system for task allocation and an approach for generating a behavior tree creating the autonomous behavior using a library of available behaviors will be established. The validity of the proposed approach is demonstrated in simulation and with multiple experiments in a laboratory environment. The result does show that the approach promisingly solves the stated problems.

In industrial robotics, the detection of failures is a key part of the robotsprogram to reach a robust assembly process. However, the setting up of sucha detection is long, very specific to a robotic operation and involves programmingby engineers. In response to this problematic, the thesis presented inthis paper proposes an algorithm which makes it generic and semi-automaticthe creation of a failures detection and a classification method. Based onmachine learning, the algorithm is able to learn how to differentiate betweena success and a failure scenario given a series of examples. Moreover, theproposed method makes the teaching of failure detection/classification accessiblefor any operator without any special programming skills.After the programming of movements for a standard behavior, a trainingset of sensory acquisitions is recorded while the robot performs blindlyoperation cycles. Depending on sensors nature, the gathered signals canbe binary inputs, images, sounds or other information measured by specificsensors (force, enlightening, temperature...). These signals contain specificpatterns or signatures for success and failures. The set of training examples isthen analyzed by the clustering algorithm OPTICS. The algorithm providesan intuitive representation based on similarities between signals which helpsan operator to find the patterns to differenciate success and failure situations.The labels extracted from this analysis are thus taken into account to learna classification function. This last function is able to classify efficiently anew signal between the success and failure cases encountered in the trainingperiod and then to provide a relevant feedback to the robot program. Arecovery can then easily be defined by an operator to fix the flaw.

Integration of renewable resources such as remote solar or wind farms and electricpower trading between neighbouring countries lead to new requirements on the development of thetransmission grids. Since AC grid expansion is limited by e.g. legislations issues, High VoltageDirect Current (HVDC) technology with its diverse benets compared to AC is being considered asappropriate alternative solution. The developed HVDC grid can be either embedded inside one ACgrid or connects several AC areas. In both architectures, the separate DC supervisory control can beproposed to control the HVDC grids using the interfacing information from AC Supervisory ControlAnd Data Acquisition (SCADA). The supervisory control is supposed to calculate the optimal power ow (OPF) in order to run the system in the most optimal situation. Based on the architecture, therequired information, boundary of the system and also objective function can vary. The aim of the thesis is to present the ndings of a feasibility study to implement a supervisorycontrol for bipolar Voltage Source Converter (VSC) HVDC grids in possible real time platforms. DCsupervisory control has a network topology manager to identify the grid conguration and employsan OPF calculator based on interior point optimization method to determine the set-point valuesfor all HVDC stations in a grid. OPF calculator takes into account the DC voltage, converter andDC line constraints.ii

Language grounding is the linking of a concept or object with its representation.This thesis tries to answer the question of whether stateof-the-art solutions for object detection and speech recognition are sufficientfor a simple language grounding system implemented on a Pepperrobot. To answer this question, three different speech recognitionand three different language grounding implementations were tested incustom datasets. The solutions for speech recognition and languagegrounding that fit better the system needs were chosen and they wereimplemented in the robot Pepper to conduct an experiment in which anobject was referred to by describing it out loud. This was done 19 timeswith different queries. The results obtained from the speech recognitionshow that Vosk is the best implementation of the tested software havingan average word error rate of 8.37%. For the language grounding,the best results were obtained from the MAttNet model, providing up toa 75.00% accuracy. Finally, the language grounding system using Pepperwas able to identify 52.63% of the queries. The conclusion is that thestate-of-the-art solutions are sufficient for a simple language groundingsystem implemented on a Pepper robot, but the implementation used inthis thesis may be improved as it is explained in the last section of thisdocument. Also, future work is proposed.

Collaborative robots are a new type of robot that unlike industrial robots are to be used in close relation to humans without physical walls. Due to this, high demands are placed on the safety of collaborative robots as people must not be harmed. To verify the safety between robot and human, collision measurements are made. For the past 10 years, the number of published articles dealing with Human RobotCollaboration has gone from almost none to approximately 3000 published articles per year. It can also be seen that research on collisions together with cooperation between robots and humans has increased during the same period but not at all in the same size. The thesis is made together with Scania which is a truck manufacturer in Sweden. Scania has its own requirements for safety around collaborative robots and they require that collision measurements in the form of pressure and force take place during each implementation of a collaborative robot. The purpose is to provide a broader knowledge of collaborative robots and their safety. To provide knowledge of how collision measurements should be made and with which tools. Using a spiral development model, several practical measurements were performed on a collaborative robot. Other methods used were interviews, literature study, calculations and benchmarking. The results show that the values set in the robot do not always correspond to those obtained during practical feeds. It uses the robot to show permissible forces in a collaborative position, but also many impermissible force measurements have taken place at different switches and at different speeds. The distance between the robot's base and the robot’s arm where the collision cuts make a great impact on the measurements and show that it can be a danger. Running a collaborative robot at high speeds without external safety equipment also demonstrates risks. Pressure measurements show a large uncertainty of the result and the result is also above the permissible limit in several cases. The impermissible pressures indicate danger for people to work in close contact with collaborative robots. The high difference in the measurement result is only for pressure measurements and the result at 100measurements over force gave a low variation.