This research aims to develop an autonomous social robot for elderly individuals. The robot will learn from the interaction and change its behaviors in order to enhance the interaction and improve the user experience. For this purpose, we aim to use Deep Reinforcement Learning. The robot will observe the user’s verbal and nonverbal social cues by using its camera and microphone, the reward will be positive valence and engagement of the user.

Safety in human-robot interaction can be divided into physical safety and perceived safety, where the later is still under-addressed in the literature. Investigating perceived safety in human-robot interaction requires a multidisciplinary perspective. Indeed, perceived safety is often considered as being associated with several common factors studied in other disciplines, i.e., comfort, predictability, sense of control, and trust. In this paper, we investigated the relationship between these factors and perceived safety in human-robot interaction using subjective and objective measures. We conducted a two-by-five mixed-subjects design experiment. There were two between-subjects conditions: the faulty robot was experienced at the beginning or the end of the interaction. The five within-subjects conditions correspond to (1) baseline, and the manipulations of robot behaviors to stimulate: (2) discomfort, (3) decreased perceived safety, (4) decreased sense of control and (5) distrust. The idea of triggering a deprivation of these factors was motivated by the definition of safety in the literature where safety is often defined by the absence of it. Twenty-seven young adult participants took part in the experiments. Participants were asked to answer questionnaires that measure the manipulated factors after within-subjects conditions. Besides questionnaire data, we collected objective measures such as videos and physiological data. The questionnaire results show a correlation between comfort, sense of control, trust, and perceived safety. Since these factors are the main factors that influence perceived safety, they should be considered in human-robot interaction design decisions. We also discuss the effect of individual human characteristics (such as personality and gender) that they could be predictors of perceived safety. We used the physiological signal data and facial affect from videos for estimating perceived safety where participants’ subjective ratings were utilized as labels. The data from objective measures revealed that the prediction rate was higher from physiological signal data. This paper can play an important role in the goal of better understanding perceived safety in human-robot interaction.

This article surveys reinforcement learning approaches in social robotics. Reinforcement learning is a framework for decision-making problems in which an agent interacts through trial-and-error with its environment to discover an optimal behavior. Since interaction is a key component in both reinforcement learning and social robotics, it can be a well-suited approach for real-world interactions with physically embodied social robots. The scope of the paper is focused particularly on studies that include social physical robots and real-world human-robot interactions with users. We present a thorough analysis of reinforcement learning approaches in social robotics. In addition to a survey, we categorize existent reinforcement learning approaches based on the used method and the design of the reward mechanisms. Moreover, since communication capability is a prominent feature of social robots, we discuss and group the papers based on the communication medium used for reward formulation. Considering the importance of designing the reward function, we also provide a categorization of the papers based on the nature of the reward. This categorization includes three major themes: interactive reinforcement learning, intrinsically motivated methods, and task performance-driven methods. The benefits and challenges of reinforcement learning in social robotics, evaluation methods of the papers regarding whether or not they use subjective and algorithmic measures, a discussion in the view of real-world reinforcement learning challenges and proposed solutions, the points that remain to be explored, including the approaches that have thus far received less attention is also given in the paper. Thus, this paper aims to become a starting point for researchers interested in using and applying reinforcement learning methods in this particular research field.

Purpose: The purpose of this thesis is to examine the market conditions for automation providers in Sweden and address the skills shortage associated with the expansion of automated production. The study aims to explore how automation providers can support small and medium-sized enterprises (SMEs) in overcoming the skills gap to enhance their adoption of advanced automation technologies.

Method: This research employs an exploratory qualitative approach, utilizing multiple-case studies. Data collection is conducted through in-depth interviews with industry experts and representatives from various companies in the Swedish manufacturing sector.

Findings: The findings reveal significant opportunities and challenges in the Swedish automation market. Key challenges include the skills shortage and the high demand for automation solutions among SMEs. The study identifies that tailored training programs, comprehensive support services, and collaborative initiatives with educational institutions are crucial strategies for addressing these challenges. Additionally, the adoption of user-friendly technologies can help SMEs integrate automation more effectively.

Practical Implications: The insights from this study can be used by automation providers to develop strategies that support SMEs in bridging the skills gap. Practical implications include the need for targeted training programs, enhanced support services, and strategic partnerships with educational institutions to ensure a skilled workforce capable of handling advanced automation technologies.

Theoretical Implications: This study contributes to the existing body of knowledge by providing a detailed analysis of the skills gap in the automation sector and proposing actionable strategies to address it. The research highlights the importance of workforce development in the successful adoption of automation technologies.

Delimitations: This thesis is delimited to the Swedish manufacturing sector, focusing on small and medium-sized enterprises (SMEs) and automation providers. The scope includes examining the current market conditions, skills gap, and potential strategies for addressing these challenges.

An Experimental Diagnostic Tool (EDT) was developed to increase the efficiency of the diagnostic process of robotic lawn mowers which resulted in a 200% productivity increase when utilizing a proposed formula specifically designed for the diagnostic process of robotic lawn mowers. The tool does not only extract and process data from the robotic lawn mower wirelessly but also highlights potential faults through an intuitive and easy-to-use interface - empowering servicing technicians who perform the diagnostics to perform at a higher level.

Further, efficiency is a term widely used in various domains and contexts when measuring the capacity of a process. Proposed general definitions of the term have been given by previous authors and researchers. However, due to a lack of universally set definitions which fit all situations, the term remains ambiguous when improvements to a specific process are needed. The unclear definition is due to the variations within each process affecting the definition of both the term itself, but also similar terms fundamentally connected to it such as productivity, performance, and profitability. The following report contains an investigation of exploratory research where the understanding of efficiency and its related concepts are analyzed within the after-sales diagnostic process of robotic lawn mowers.

As manufacturing industry seeks different strategies and technologies to respond to the ever-increasing demands in markets that prioritize versatility of products with low-volume productions, certain technologies and strategies gain more attraction and form higher acceptance levels among different sectors. Individual firms are driven by their market requirements. Various factors including product specification, assembly sequence, and manufacturing operations are central to the decisions that are made with respect to the type of technology to respond to market dynamics. Additive Manufacturing (AM) is one of the technology alternatives that has exhibited remarkable strengths in countering market disruptions. Although AM can be utilized along conventional technologies (i.e., subtracting and forming) in a hybrid context to combine advantages and offset weaknesses of each category, the arguments supporting its applications would need to be formulated rigorously to ensure investments are rightfully justified. Another alternative continuously investigated by companies is automation and more specifically, using robotics for various purposes e.g., operations like welding and painting, material handling, machine tending, etc. Both industrial robots and the applications that require a collaboration between humans and robots can be valid in this context. Considering advancements in AM and Automation and their potentials in increasing flexibility, expediting operations, and leveraging cost advantages, this paper explores how AM and automation in tandem could improve flexibility in productions. Results of this study can be used for proposing a conceptual model which will be further developed and then tested on industrial cases in future studies. While this study incorporates raw data about processing requirements in production that has been obtained via interviews with industrial companies, inputs about the technologies i.e., AM and robotics are derived from literature.

I takt med att marknaden för robotgräsklippare ökar så är pressen högre på företag att deras produkt ska vara robust. Detta kan uppnås genom att testerna som görs på robotgräsklipparen testas så fort som möjligt. Genom att skapa en simulering där alla tester genomförs istället för att köra testerna på en fysisk robotgräsklippare kan detta uppnås och utifrån detta utformades forskningsfrågan. För att skapa simuleringen undersöktes först mjukvaror vilket ansågs lämpliga för att utveckla en simulator, detta gjordes via en fallstudie. Dessa har sedan analyserats och jämförts för att till slut bestämma de som ansetts bäst att använda. Med hjälp av de så startade en utveckling av en simulator där hjul- och kollisionsdata hämtades från en fysisk robotgräsklippare och skickades till en virtuell robotgräsklippare. Den färdigställda simulatorn utvärderades vid slutet av arbetet med hjälp av experiment där författarna observerade och jämförde rörelsen hos den fysiska och virtuella robotgräsklipparen. För att utföra denna uppgift så tillämpade arbetet metoden Design science research där det arbetades iterativt vid utvecklingen av simulatorn. Resultatet visar på att det är möjligt att skapa en simulator med de valda mjukvarorna ROS och Gazebo där man kan genomföra simulerade tester. Arbetet visar på ökad kunskap där data från en fysisk robotgräsklippare kan implementeras i 3D-simulatorn Gazebo via ramverket ROS. Studien kan användas som riktlinje i liknande projekt när det kommer till val av mjukvaror och om de är lämpliga. Arbetet begränsas till enbart hjul- och kollisionsdata från den fysiska robotgräsklipparen.

Autonomous robots are becoming a bigger part of our society. This thesis aims to evaluate a robot testing dashboard (RTD) that can be used as a new way of finding improvements when developing autonomous robots that do not use machine learning.

The method that is used is design science research, which is used when creating and evaluating an artifact to address a practical problem. In our case the artifact isthe RTD.

This project was performed at a company called Greenworks, which among other things develops and sells autonomous lawn mowers. The company wants to find new testing methods to help develop their autonomous lawnmowers. The RTD is created to visualize the inputs that the lawn mower utilizes to perform its tasks. A human pilot will then control the lawn mower, by only looking at that visualized data. If the pilot using the RTD can execute the same tasks as the lawn mower in its autonomous mode, the test results can be analyzed to see whether the human has done some parts of the tasks differently. The best outcome from the analysis of the test results is to find areas of improvement that can be implemented into the autonomous lawn mower design, both in software and hardware.

For this purpose, an RTD was built and tested at Greenworks. From the tests using the RTD we concluded that it is helpful in the testing process, and we could find areas of improvements after analysis of our tests. However, the use of the RTD will require more time and resources compared to other methods. Each company that uses a similar dashboard concept will have to evaluate if the benefits are worth the time. Furthermore, the concept may not suit all areas of robotics but does seem to suit situations where a human can have an advantage over robots, such as in creative problem solving.

Reinforcement learning algorithms have proven to be capable of solving complicated robotics tasks in an end-to-end fashion without any need for hand-crafted features or policies. Especially in the context of robotics, in which the cost of real-world data is usually extremely high, Reinforcement Learning solutions achieving high sample efficiency are needed. In this paper, we propose a framework combining the learning of a low-dimensional state representation, from high-dimensional observations coming from the robot's raw sensory readings, with the learning of the optimal policy, given the learned state representation. We evaluate our framework in the context of mobile robot navigation in the case of continuous state and action spaces. Moreover, we study the problem of transferring what learned in the simulated virtual environment to the real robot without further retraining using real-world data in the presence of visual and depth distractors, such as lighting changes and moving obstacles. A video of our experiments can be found at: https://youtu.be/rUdGPKr2Wuo.

Collaborative swarms of robots/UAVs constitute a promising solution for precision agriculture and for automatizing agricultural processes. Since agricultural fields have complex topologies and different constraints, the problem of optimized path routing of these swarms is important to be tackled. Hence, this paper deals with the problem of optimizing path routing for a swarm of ground robots and UAVs in different popular topologies of agricultural fields. Four algorithms (Nearest Neighbour based on K-means clustering, Christofides, Ant Colony Optimisation and Bellman-Held-Karp) are applied on various farm types commonly found around Europe. The results indicate that the problem of path planning and the corresponding algorithm to use, are sensitive to the field topology and to the number of agents in the swarm.

Reinforcement Learning has been able to solve many complicated robotics tasks without any need for feature engineering in an end-to-end fashion. However, learning the optimal policy directly from the sensory inputs, i.e the observations, often requires processing and storage of a huge amount of data. In the context of robotics, the cost of data from real robotics hardware is usually very high, thus solutions that achieve high sample-efficiency are needed. We propose a method that aims at learning a mapping from the observations into a lower-dimensional state space. This mapping is learned with unsupervised learning using loss functions shaped to incorporate prior knowledge of the environment and the task. Using the samples from the state space, the optimal policy is quickly and efficiently learned. We test the method on several mobile robot navigation tasks in a simulation environment and also on a real robot. A video of our experiments can be found at: https://youtu.be/dgWxmfSv95U.

In this research, we investigate the use of Reinforcement Learning (RL) for an effective and robust solution for exploring unknown and indoor environments and reconstructing their maps. We benefit from a Simultaneous Localization and Mapping (SLAM) algorithm for real-time robot localization and mapping. Three different reward functions are compared and tested in different environments with growing complexity. The performances of the three different RL-based path planners are assessed not only on the training environments, but also on an a priori unseen environment to test the generalization properties of the policies. The results indicate that RL-based planners trained to maximize the coverage of the map are able to consistently explore and construct the maps of different indoor environments.

Autonomous robots struggle with plan adaption in uncertain and changing environments. Although modern robots can make popcorn and pancakes, they are incapable of performing such tasks in unknown settings and unable to adapt action plans if ingredients or tools are missing. Humans are continuously aware of their surroundings. For robotic agents, real-time state updating is time-consuming and other methods for failure handling are required. Taking inspiration from human cognition, we propose a plan adaption method based on event segmentation of the image-schematic states of subtasks within action descriptors. For this, we reuse action plans of the robotic architecture CRAM and ontologically model the involved objects and image-schematic states of the action descriptor cutting. Our evaluation uses a robot simulation of the task of cutting bread and demonstrates that the system can reason about possible solutions to unexpected failures regarding tool use.

Optimising object order in stacking problems remains a hard problem for cognitive robotics research. In this paper, we continue our work on using the spatiotemporal relationships called image schemas to represent affordance spaces founded on object properties. Based on object properties, we introduce a stacking-order algorithm and describe the action descriptors using an image-schematic event segmentation format by describing a small subset using the Image Schema Logic ISL^𝐹𝑂𝐿.

In this extended abstract, we present initial work on intelligent object stacking by household robots using a symbolic approach grounded in image schema research. Image schemas represent spatiotemporal relationships that capture objects’ affordances and dispositions. Therefore, they offer the first step to ground semantic information in symbolic descriptions. We hypothesise that for a robot to successfully stack objects of different dispositions, these relationships can be used to more intelligently identify both task constraints and relevant event segments. 

The potential of Robotic Process Automation (RPA) for enterprises is undisputed and includes automation of routine tasks, improvement of data quality or reduction of monotonous tasks. However, our observation from projects with enterprises is that much of this potential is clearly visible in larger enterprises but not so obvious in small enterprises. SMEs often show a lower readiness to invest time and resources in new technologies. Thus, application fields requiring an investment are not exploited by SMEs. So, what is the real application potential of RPA in SMEs? Are there typical application scenarios with potentially high benefits that SMEs should focus on? This is the focus of this paper. We will address the question of application potential primarily from a qualitative perspective, i.e. the aim is not to quantify the potential but to qualify the application fields. The contributions of this paper are (1) an innovative dataspace as an example case for business model development, (2) an approach to support business model development based on data value chains and reference enterprise architectures, and (3) analysis of existing literature in the field. © 2020 Copyright for this paper by its authors.

Examensarbetet utfördes av två ingenjörsstudenter från Jönköpings tekniska högskola våren 2024. Syftet med undersökningen är att jämföra digitaliserad utsättning med manuell utsättning för att se om momentet går att effektivisera. Utsättning för innerväggar har genomförts på samma sätt i flera år, i jämförelse med andra områden inom byggbranschen som har utvecklas mer mot digitalisering. För att utveckla området krävs en omställning mot en mer digitaliserad riktning. Varje steg framåt mot digitalisering för byggbranschen ser författarna som ett steg framåt mot utveckling.

Information togs fram med hjälp av litteraturgenomgång samt kvalitativa semistrukturerade intervjuer som undersökningsmetoder. Varje respondent är sakkunnig inom området för att stärka validiteten av arbetet. Två grupper skapades under intervjuerna, en grupp för digitaliserad utsättning och en grupp för manuell utsättning.

Att byggbranschen ska utvecklas för det bättre är något alla strävar efter. Utsättningsmomentet för ett projekt är påfrestande och speciellt sett till ergonomi, större omfattning av projektet medför till mer utsättning vilket bidrar till mer påfrestning för arbetarna. Att yrkesarbetare arbetar hårt och gör påfrestande arbeten är inget nytt. Genom att minska antalet påfrestande arbeten för arbetarna kommer det göra skillnader för aspekterna tid, ergonomi och ekonomi. 

Målet med undersökningen är att jämföra robotutsättning med manuell utsättning och undersöka om robotutsättning är en relevant metod att använda inom momentet. Författarna har riktat in sig för utsättningsprocessen för innerväggar. Detta är för att förbättra arbetsmiljön för yrkesarbetare och för att bespara tid i projektet. Byggprojekt i dagsläget (2024) har tidsbrist och att bespara tid i varje moment kan bli avgörande för att färdigställa projektet inom tidsramen.

Studien har riktat in sig på en utsättningsrobot som är en metod som inte används lika mycket i Sverige som andra länder på grund av att det är en ny teknologi. Metoden visades sig ha brister i vissa områden men vid rätta förhållanden så kommer roboten utföra arbetet snabbare än människan, samt så är det bevisat att det är mer ergonomiskt än manuell utsättning. Intervjumaterialen visade på att de som har testat på robotutsättning är mer positiva mot produkten jämfört med respondenterna som endast gjort manuell utsättning. Undersökningsmetoderna gav underlaget för att besvara frågeställningarna för studien.

In order for humans and robots to collaborate on an assembly line, safety of operations is a prerequisite. In this article, two assembly stations where a large industrial robots collaborate with humans will be analysed with the aim to 1. determine the characteristics of hazards associated with human-robot interaction and 2. design solutions that can mitigate risks associated with these hazards. To support the aim of this article, a literature review will attempt to characterize automation and detail the problems associated with human-automation interaction. The analysis points at situational awareness and mode-awareness as contributing factors to operator and process safety. These underlying mechanisms, if recognised by the risk assessment team as hazards, can mitigate risks of operator injury or production delays. This article details the function of visual and physical interfaces that allow operators to comprehend system-state in order to avoid undesirable situations. 

Several real-world applications introduce derivativefree optimization problems, called variable dimension problems, where the problem's dimension is not known in advance. Despite their importance, no unified framework for developing, comparing, and benchmarking variable dimension problems exists. The robot arm controlling problem is a variable dimension problem where the number of joints to optimize defines the problem's dimension. For a holistic study of global optimization methods, we studied 14 representative methods from 4 different categories, i.e., (i) local search optimization techniques with random restarts, (ii) state-of-the-art DIRECT-type methods, (iii) established Evolutionary Computation approaches, and (iv) state-of-the-art Evolutionary Computation approaches. To investigate the effect of the problem's dimensionality on the solution we generated 20 instances of various combinations among the number of predefined and open decision variables, and we performed experiments for various computational budgets. The results attest that the robot arm controlling problem provides a proper benchmark for variable dimensions. Furthermore, methods in-corporating local search techniques have dominant performance for higher dimensionalities of the problem, while state-of-the-art EC methods dominate in the lower dimensionalities.

Denna studie handlar om robotarna Blue-bot och Bee-bot. Syftet är att ta reda på hur robotar används av utbildade pedagoger samt vilka färdigheter och kunskaper de beskriver att barnen utvecklar av att använda robotar. Frågeställningarna studien har är: “Hur beskriver de utbildade pedagogerna att de undervisar med Blue-bot eller Bee-bot?” och “Vilka färdigheter och kunskaper beskriver de utbildade pedagogerna att barnen utvecklar av att använda Blue-bot eller Bee-bot?”. Datainsamlingen gjordes via en kvalitativ metod med en fokusgrupp och tre intervjuer samt utfördes digitalt via Teams. Studien använder sig av den sociokulturella teorin för att granska resultatet. Resultatet visar att deltagarnas huvudsakliga syfte med arbetet med robotarna är att arbeta med matematik, vilket även involverar problemlösning, uppskattning av avstånd samt öva på rumsuppfattning och sekvensering. Alla delar inom matematik är färdigheter Kiliç (2022) kopplar till datalogiskt tänkande och det kan därmed sägas att det deltagarna vill att barnen ska utveckla i användandet av robotarna är datalogiskt tänkande. Resultatet visar också på att om syftet med att använda robotarna är att lära barnen programmeringsbegrepp och matematiskt tänkande är den analoga programmeringen bättre att använda än robotarna. Resultatet visar även att pedagogers frånvaro kan ge barnen större möjlighet till att utveckla det sociala samspelet samt utveckla fler färdigheter genom stöttning av andra barn.

Purpose – Using the problem description, the following purpose has been formulated:

The purpose of the work is to identify the possibility, as well as profitability, of increased efficiency in internal transport through AGV-solutions in manufacturing companies.

Based on the purpose, the following research questions have been formulated:

I. What requirements are set on companies to increase efficiency of internal transports through AGV-solutions?

II. How does the introduction of AGV-solutions affect the efficiency of internal transports?

III. What factors can impact the calculations of an AGV-solutions’ profitability?

Method – The study is a case study in a single-case study direction, where the data has been collected using observations, interviews, frequency studies and document studies to answer the study’s research questions. Furthermore, the methods are mainly based on the collection of data at the case company, but also with the truck suppliers. As the work is based on one case, many theories have been studied to create a generalizable work. Finally, a lot of focus during the work has been on securing a high level of credibility in the work. Examples of this are triangulation of methods, use of peer-reviewed sources and transparency about the execution.

Findings – The result of the study highlights a clear need for businesses to analyze their workflows when considering the implementation of an AGV solution. This should be done to ensure a more efficient implementation with more promising outcomes. Furthermore, the results have shown that the adoption of AGV solutions is most relevant and applicable in simple, repetitive workflows. Finally, it has been identified that the profitability of implementation can vary as businesses requirements differ. However, the implementation creates opportunities for staff redeployment, which can contribute to increased profitability.

Implications – Further implications of the work touch upon the importance of conducting a current situation analysis. Additional implications include the use of standardization to ensure smooth operations, and investment calculations to determine profitability. However, precise instructions on how the methods and theories should be applied are not feasible. This means that businesses themselves need to adapt these according to their specific needs, implying that the study should be seen as a template rather than a manual.

Limitations – Since the study is a single-case study, it becomes more difficult to generalize the results. More time in combination with more data collected, would have given the study stronger results and credibility.

One of the problems an agent faces when operating in a partially known, dynamic, sometimes unpredictable environment is to keep track of aspects of the world relevant to its task, and, if possible, restrict its attention to only these aspects. We present our first steps towards constructing a system that combines image schematic knowledge and reasoning with reactive robotics, and which enables perception that focuses on, and keeps track of, relevant entities and relationships. While our approach is more reasoning intensive than is usual in reactive robotics, the formalism we use for inference is fast and allows an agent to adjust, in real time, the complexity of its action selection procedures according to the complexity of the relevant part of the environment. We illustrate our approach with a few simulated examples of robots performing navigation tasks. In some examples, interaction with obstacles is necessary to complete the navigation tasks, adding complexity to the scenario.

When designing social robots, it is crucial to understand the diverse expectations of different kinds of innovation adopters. Different factors influence early adopters of innovations and mass market representatives’ perceptions of the usefulness of social robots. The first aim of the study was to test how applicable the technology acceptance model 3 (TAM3) is in the context of social robots. Participants’ acceptance of social robotics in a workplace environment in the fuzzy front-end (FFE) innovation phase of a robot development project was examined. Based on the findings for the model, we developed a reduced version of the TAM3 that is more applicable for social robots. The second objective was to analyze how early adopters’ and mass market representatives’ acceptance of social robots differs. Quantitative research methods were used. For early adopters, result demonstrability has a significant influence on perceived usefulness of social robots, while for mass market representatives, perceived enjoyment has a more significant influence on perceived usefulness. The findings indicate that users’ innovation adoption style influences the factors that users consider important in the usefulness of social robots. Robot developers should take these into account during the FFE innovation phase.

A stagnation of productivity increase has been observed in the Swedish wood product industry. The manufacturers believe that there is a need to invest further in automation in order to stay competitive. For this reason, the paper seeks to understand the role of automation in the Swedish wood product industry, and focuses on identifying the challenges and opportunities of automation. The following research question was addressed: What are the challenges and opportunities of automation in the wood product industry? In order to answer the research question, four case studies were conducted, each case representing a different business area. The research question was examined in terms of internal – and external challenges and opportunities. The internal challenges and opportunities examine the manufacturing, while the external challenges and opportunities examine the influence of the business environment. Findings indicate that lack of manufacturing strategies, and lack of awareness of automation technologies were some of the main challenges. Regarding the opportunities, increased profitability and competitiveness were emphasized. The identification of the challenges and opportunities of automation in the wood product industry can provide insights and be used as underlying decisions for automation investments.

The role of warehouse in the value stream, plays a significant part in how well a company can satisfy the requirements of a customer. To be able to support the demand of the customers, production companies are forced to produce a bigger variety of models in the product range, which results in an increasing assortment. With an effective stock keeping, there is a potential to reduce waste and increase profitability. In the choice of warehouse-design, warehouse-operations have the opportunity of choosing between a number of different designs containing manual and automated put away.It has become common that companies are choosing a combination of automated and manual processes. With this combination of processes, the companies are faced with choices about stock locations. With the help of well developed criterias about stock locations, there is a potential to achieve a more effective put away. The purpose of the study has been formulated as ”Investigate when automation in warehouse management makes the put away process in a central warehouse more efficient”.

Through a case study on a warehouse, located in Nässjö, this report will answer the research questions regarding which of the manual and automated put away processes to use. The research questions are:

1. Which criteria affect the decision of automated or manual put away?
2. Based on the criteria that are raised in research question 1, when is it more efficient to store carton in AS/RS in comparison with manual picking with truck in pallet racking?

The research questions wish to lift criteria as a basis for decision making and when it is more effective to use automated put away. The warehouse that this study is based on is using both automated and manual materials handling. The specific automation system that has been the basis for this study’s result is AutoStore. This report has a qualitative approach, the methods that has been used for collecting data are interviews, observations and document studies.

Dimension, process time and utilization were the three criteria that are presented in the result. Based on the pallets limit of space in the automated system, the dimension of the article is the first criteria. In addition to the article’s measurements, the articles shape and material of the package had an impact if the article would fit in the load carrier. The two put away processes include different operations that differ in how long they take to execute, which has an impact on the company's efficiency. One reason for the time consuming difference, turned out to be how the quantity affects the time for put away in AutoStore comparing with the constant time for manual put away. A time consuming operation in the process with AutoStore was the one with repacking for the systems pallet. The operation also turned out to affect the utilization rate, due to its measurement of a pallets fill rate.  The quantity of put away in combination with the dimensions of the article, affects the total utilization rate.

With the result, the aim of the study is to contribute to companies that contains a combination of warehouse-design with a basis for the decision making about stock locations. 

Purpose – The purpose of the thesis is to evaluate the accuracy of two different approaches of how to detect skidding in autonomous lawn mowers. Using the motor currents and inertia sensors from the mower, a neural network is applied to predict if there was a skidding or not. The usefulness of knowing when a skid happens could be of value for future developments making better autonomous decision making.

Method – The thesis will adopt the methodology process of Design Science Research (DSR). The study begins with bringing awareness of the problem by previous knowledge and related works to skidding in wheeled robots. Thereafter an experiment is set up to generate data when the autonomous lawn mower is in conditions of skidding and non-skidding. The data collected will be processed with machine learning algorithm, multilayer perceptron.

Findings – The findings showed high accuracies in both techniques where adding an IMU sensor in addition to motor currents showed higher accuracy then only using motor currents. Both techniques showed low number of false detections and near zero missed detections which is a preferred feature, the behavior of the autonomous lawn mower benefits more from a false detection than not detecting any at all and get stuck.

Implications – The autonomous lawn mowers of today have a tendency of failing in the yard due to skidding in uneven or slippery terrain. The robot either reacts by assuming a collision has happened or gets no traction and gets stuck. A first step to solve this problem is by detecting such a skid to then be able to take action.

Limitations – The results will be limited to the autonomous lawn mower of Globe Group as the data collection is made with an autonomous lawn mower of Globe Group. The mower will run on a flat outdoor grass lawn to maintain the experiment on a reasonable level for a bachelor thesis.

Keywords – robot, autonomous, lawn mower, machine learning, detection, skidding

Purpose – The purpose of the study was to investigate what precision a radar sensor could guide an autonomous robot into a docking station.

Method – The study was conducted with design science research and it was produced artifacts that were evaluated through experiments. The work was based on the hypothesis that two cube corner reflectors, mounted on a docking station, made it possible to locate a docking station with a radar sensor. The study was based on that the robot had navigated in another way to a circular area, with a radius of 0.5 meters, located 1.5 meters straight in front of the docking station. The initial experiments were designed to investigate how different locations of the reflectors affected the radar sensor's ability to detect and position the docking station from the area just mentioned. The results from these experiments was then used to design the final experiment of the study. In this a robot was equipped with a radar sensor. Two reflectors, mounted on a common stand, was used to imitate a docking station. The experiment evaluated how the location of the docking station was affected by the radar's movement and what precision a docking could be carried out with the concept.

Findings – The study has shown that there are times when two cube corner reflectors mounted on a docking station can be used to guide a robot, equipped with a radar sensor, into a docking station. In the final experiment of the study, ten dockings were carried out with the concept. The largest angle deviation measured, at the robot's final position in the docking station, was three degrees and the largest lateral offset was six millimetres. The analyses of the data collected during the experiment showed that there was potential to reduce these differences.

The study also showed that the ability of the radar system to locate the docking station increased when the reflectors were at different distances from the radar compared to if they were only at different angles seen from the radar.

Implications – The study contributes to an increased knowledge of how a radar sensor can be used to guide an autonomous robot into a docking station. The study also provides suggestions for further research that can add additional knowledge.

Limitations – A docking procedure depends on a variety of parameters such as, the environment around the docking station, which sensors are available on the robot and what the requirements are on the precision of the docking. This study has shown that a radar guided docking may be carried out for at least one combination of parameters. It has not investigated all possible limiting factors that may exist. One of these factors could be the radar's ability to distinguish a docking station in an environment with many other objects.

PurposeThe purpose of the thesis is to use design science research to produce information to improve a support vector machine (SVM) model and to gain knowledge about optimization of hyperparameters through experiments. The objective will be answered through evaluation of performance gained by adjusting the hyperparameters. The report will also evaluate the further use of the techniques with IR cameras in combination with histogram of oriented gradients (HOG) as translated input for the SVM-model trained to detect and do a classification of runways.MethodThe method was performed through three technical experiments performed at Saab Avionics Systems. The experiments contain adjusting and optimizing settings in Saabs framework for detecting runways with the machine learning method support vector machine. The experiments with manual grid search will test and evaluate the performance of the support vector ratio, accuracy and more. Three different optimization methods will be evaluated which consists of grid search, random grid search and Bayesian optimization and all three will optimize the parameters C and gamma.FindingsThe findings indicate that all the evaluated parameters could bring positive performance impact on the SVM. Some important results were the HOG settings 6 orientation histograms, 64*64 pixels per cell, 2*2 cells per pixel, which performed best. An even dividable resolution of the image was also important for the HOG calculations. Other results indicate that the Radial Basis Function - kernel was the better choice over the linear kernel, as well as a higher amount of training data was better for a higher performance.The best performing optimization algorithm was the Bayesian optimization which found and improved three of eight testcases. The technique combining SVM with HOG-vectors for detecting runways are shown being implementable with high performance. A hardware implementation with an FPGA would be the recommended implementation which would improve classification time as well as the HOG computation time. Around 35-145 fps per detection window would be achievable with this technique with an accuracy above 99.4%.ComplicationsFew implications were found during the work apart from one. Which were the case of the detection window of one of the IR cameras (LWIR) which was a bit skewed compared to the SWIR sensor. The result contains some uncertainties of the performance compared between SWIR and LWIR sensors.LimitationsThe limitation of the report is that the training data were similar and only consisting of one runway. The accuracy performance can be a bit misleading when evaluating an imbalanced class.KeywordsSVM, HOG, runway, hyperparameters, optimization, RBF kernel, grid search, Bayesian optimization, random grid search, object detection, classification, margin for detecting window, over fitting, C, gamma, range, IR camera, training data, data augmentation, class balance in data, implementation.