Self-balancing robots are unique mobile platforms that stay upright on two wheels using a closed-loop control system. They can turn on the spot using differential steering and have compact form factors that limit their required floor space. However they have major limitations keeping them from being used in real world applications: they cannot stand-up on their own, climb stairs, or overcome obstacles. They can fall easily if hit or going onto a slippery surface because they rely on friction for balancing. The first part of this research proposes a novel design to address the above mentioned issues related to stair-climbing, standing-up, and obstacles. A single revolute joint is added to the centre of a four-wheel drive robot onto which an arm is attached, allowing the robot to successfully climb stairs and stand-up on its own from a single motion. A model and simulation of the balancing and stair-climbing process are derived, and compared against experimental results with a custom robot prototype. The second part, a control system for an inverted pendulum equipped with a gyroscopic mechanism, was investigated for integration into self-balancing robots. It improves disturbance rejection during balance, and keeps equilibrium on slippery surfaces. The model of a gyroscope mounted onto an actuated gimbal was derived and simulated. To prove the concept worked, a custom-built platform showed it is possible for a balancing robot to stay upright with zero traction under the wheels.

A mobile robot needs to know its pose (position and orientation) in order to navigate and perform useful tasks. The problem of determining this pose with respect to a global or local frame is called localisation, and is a key component in providing autonomy to mobile robots. Thus, localisation answers the question Where am I? from the robot’s perspective. Localisation involving a single robot is a widely explored and documented problem in mobile robotics. The basic idea behind most documented localisation techniques involves the optimum combination of noisy and uncertain information that comes from various robot’s sensors. However, many complex robotic applications require multiple robots to work together and share information among themselves in order to successfully and efficiently accomplish certain tasks. This leads to research in collaborative localisation involving multiple robots. Several studies have shown that when multiple robots collaboratively localise themselves, the resulting accuracy in their estimated positions and orientations outperforms that of a single robot, especially in scenarios where robots do not have access to information about their surrounding environment. This thesis presents the main theme of most of the existing collaborative, multi-robot localisation solutions, and proposes an alternative or complementary solution to some of the existing challenges in multirobot localisation. Specifically, in this thesis, a heuristically tuned Extended Kalman Filter is proposed to localise a group of mobile robots. Simulations show that when certain conditions are met, the proposed tuning method significantly improves the accuracy and reliability of poses estimated by the Extended Kalman Filter. Real world experiments performed on custom-made robotic platforms validate the simulation results.

In Wireless Sensor and Robot Networks (WSRNs), static sensors report event information to one of the robots. In the k nearest neighbour query processing problem in WSRNs, the robot receives event report needs to find exact k nearest robots (KNN) to react to the event, among those connected to it. We are interested in localized solutions, which avoid message flooding to the whole network. Several existing methods restrict the search within a predetermined boundary. Some network density-based estimation algorithms were proposed but they either result in large message transmission or require the density information of the whole network in advance which is complex to implement and lacks robustness. Algorithms with tree structures lead to the excessive energy consumption and large latency caused by structural construction. Itinerary based approaches generate large latency or unsatisfactory accuracy. In this thesis, we propose a new method to estimate a search boundary, which is a circle centred at the query point. Two algorithms are presented to disseminate the message to robots of interest and aggregate their data (e.g. the distance to query point). Multiple Auction Aggregation (MAA) is an algorithm based on auction protocol, with multiple copies of query message being disseminated into the network to get the best bidding from each robot. Partial Depth First Search (PDFS) attempts to traverse all the robots of interest with a query message to gather the data by depth first search. This thesis also optimizes a traditional itinerary-based KNN query processing method called IKNN and compares this algorithm with our proposed MAA and PDFS algorithms. The experimental results followed indicate that the overall performance of MAA and PDFS outweighs IKNN in WSRNs.

In this thesis fundamental pieces of multi robot systems have been discussed and researched, in order to develop and build a system with easily obtainable electronics and to answer how much communication is needed as well as which design choices are important to make it robust. Our work will hopefully contribute to others in the community who are working with Raspberry Pi and Windows 10 IoT Core by being open source. As a result a proof of concept system of two simple robots has been implemented. In this paper we have used trigonometry and dead-reckoning for localization, when coordinating our robots a leader/follower model has been applied. The communication has been developed using the AllJoyn framework to develop an interface that has IoT capabilities. The results show that our system has working communication and simulated localization, however the limitations in the hardware results in an error in localization which we present in this paper. To answer our research questions the amount of communication needed is dependent on the problem and how many robots you need to apply in order to solve it and the most important design choice for current multi robot systems is a controlled environment.

This dissertation presents the results of a series of studies on dynamic stabilisation of CoMan, which is actuated by series elastic actuators. The main goal of this dissertation is to dynamically stabilise the humanoid robot on the floor by the simplest multivariate feedback control for the purpose of walking. The multivariable scheme is chosen to take into account the joints' interactions, as well as providing a systematic way of designing the feedback system to improve the bandwidth and tracking performance of CoMan's existing PID control. A detailed model is derived which includes all the motors and joints state variables and their multibody interactions which are often ignored in the previous studies on bipedal robots in the literature. The derived dynamic model is then used to design multivariable optimal control feedback and observers with a mathematical proof for the relative stability and robustness of the closed loop system in face of model uncertainties and disturbances. In addition, two decentralized optimal feedback design algorithms are presented that explicitly take the compliant dynamics and the multibody interactions into account while providing the mathematical proof for the stability of the overall system. The purpose of the proposed decentralized control methods is to provide a systematic model based PDPID design to replace the existing PID controllers which are derived by a trial and error process. Moreover, the challenging constrained and compliant motion of the robot in double support is studied where a novel constrained feedback design is proposed which directly takes the compliance dynamics, interactions and the constraints into account to provide a closed loop feedback tracking system that drives the robot inside the constrained subspace. This method of control is particularly interesting since most control methods applied to closed kinematic chains (such as the double support phase) are over complicated for implementation purposes or have an ad-hoc approach to controller design. In terms of walking trajectory generation, an extension to the ZMP walking trajectory generation is proposed to utilise the CoMan's upper body to tackle the non-minimum phase behaviour that is faced in trajectory generation. Simple inverted pendulum models of walking are then used to study the maximum feasible walking speed and step size where parameters of CoMan are used to provide numerical upperbounds on the step size and walking speed. Use of straight knee and toe push-off during walking is shown to be beneficial for taking larger step lengths and hence achieving faster walking speeds. Subsequently, the designed tracking systems are then applied to a dynamic walking simulator which is developed during this PhD project to accurately model the compliant walking behaviour of the CoMan. A walking gait is simulated and visualized to show the effectiveness of the developed walking simulator. Moreover, the experimental results and challenges faced during the implementation of the designed tracking control systems are discussed where it is shown that the LQR feedback results in 50% less control effort and tracking errors in comparison with CoMan's existing independent PID control. This advantage directly affects the feasible walking speed. In addition, a set of standard and repeatable tests for CoMan are designed to quantify and compare the performance of various control system designs. Finally, the conclusions and future directions are pointed out.

Dans la dernière décennie, la robotique souple a connu un gain de popularité considérable. Elle est, de façon inhérente, sécuritaire pour les humains et l’environnement qui l’entourent. Grâce à sa faible rigidité, la robotique souple est idéale pour manipuler des objets fragiles et elle est en mesure de s’adapter à son environnement. Les caractéristiques uniques de la robotique souple font de cette technologie un tremplin vers la conception d’appareils médicaux novateurs, plus particulièrement pour des outils permettant le positionnement d’aiguilles dans le but de faire des interventions percutanées, notamment au niveau du foie. Toutefois, la souplesse de cette technologie induit, du même coup, quelques désagréments. Elle procure un comportement sécuritaire, mais entraîne aussi un manque de rigidité limitant les applications de la robotique souple. Sans une rigidité minimale, il est impossible d’accomplir des opérations repérables et précises. La robotique souple a en fait un compromis majeur entre la capacité de chargement et la plage d’utilisation. Pour utiliser cette technologie dans le domaine médical, il est primordial d’ajouter un système permettant de moduler la rigidité du système pour inhiber ce compromis. Couplée avec un système de freinage granulaire, la robotique souple semble comporter l’ensemble des caractéristiques permettant d’accomplir des interventions au foie. Cette étude tend à démontrer que couplée à un système modulant la rigidité, la robotique souple peut être utilisée pour accomplir des opérations d’une façon précise et repérable, tout en demeurant sécuritaire. Le positionneur d’aiguilles développé est 100 % compatible avec l’Imagerie à Résonance Magnétique (IRM). La plage d’insertion du système permet de rejoindre l’entièreté du foie (1500 cm³), tout en maintenant une rigidité suffisante (3 N/mm) et en étant aussi précis que l’outil d’imagerie utilisée (1 mm). L’approche hybride consistant à développer un système activé de façon souple couplée à un module régulant sa rigidité permet d’obtenir à la fois les avantages d’une robotique compliante (souple) et conventionnelle (dure).

abstract: Unmanned subsurface investigation technologies for the Moon are of special significance for future exploration when considering the renewed interest of the international community for this interplanetary destination. In precision agriculture, farmers demand quasi-real-time sensors and instruments with remote crop and soil detection properties to meet sustainability goals and achieve healthier and higher crop yields. Hence, there is the need for a robot that will be able to travel through the soil and conduct sampling or in-situ analysis of the subsurface materials on earth and in space. This thesis presents the design, fabrication, and characterization of a robot that can travel through the soil. The robot consists of a helical screw design coupled with a fin that acts as an anchor. The fin design is an integral part of the robot, allowing it to travel up and down the medium unaided. Experiments were performed to characterize different designs. It was concluded that the most energy-efficient speed from traveling down the medium is 20 rpm, while 60 rpm was the efficient speed for traveling up the medium. This research provides vital insight into developing subsurface robots enabling us to unearth the valuable knowledge that subsurface environment holds to help the agricultural, construction, and exploration communities. / Dissertation/Thesis / Video of Self Burrowing Robot travelling down the medium / Video of Self Burrowing Robot travelling out the medium / Masters Thesis Mechanical Engineering 2020

Dans la communauté robotique aérienne, un croissant intérêt pour les systèmes multirobot (SMR) est apparu ces dernières années. Cela a été motivé par i) les progrès technologiques, tels que de meilleures capacités de traitement à bord des robots et des performances de communication plus élevées, et ii) les résultats prometteurs du déploiement de SMR tels que l’augmentation de la zone de couverture en un minimum de temps. Le développement d’une flotte de véhicules aériens sans pilote (UAV: Unmanned Aerial Vehicle) et de véhicules aériens de petite taille (MAV: Micro Aerial Vehicle) a ouvert la voie à de nouvelles applications à grande échelle nécessitant les caractéristiques de tel système de systèmes dans des domaines tels que la sécurité, la surveillance des catastrophes et des inondations, la recherche et le sauvetage, l’inspection des infrastructures, et ainsi de suite. De telles applications nécessitent que les robots identifient leur environnement et se localisent. Ces tâches fondamentales peuvent être assurées par la mission d’exploration. Dans ce contexte, cette thèse aborde l’exploration coopérative d’un environnement inconnu en utilisant une équipe de drones avec vision intégrée. Nous avons proposé un système multi-robot où le but est de choisir des régions spécifiques de l’environnement à explorer et à cartographier simultanément par chaque robot de manière optimisée, afin de réduire le temps d’exploration et, par conséquent, la consommation d’énergie. Chaque UAV est capable d’effectuer une localisation et une cartographie simultanées (SLAM: Simultaneous Localization And Mapping) à l’aide d’un capteur visuel comme principale modalité de perception. Pour explorer les régions inconnues, les cibles – choisies parmi les points frontières situés entre les zones libres et les zones inconnues – sont assignées aux robots en considérant un compromis entre l’exploration rapide et l’obtention d’une carte détaillée. À des fins de prise de décision, les UAVs échangent habituellement une copie de leur carte locale, mais la nouveauté dans ce travail est d’échanger les points frontières de cette carte, ce qui permet d’économiser la bande passante de communication. L’un des points les plus difficiles du SMR est la communication inter-robot. Nous étudions cette partie sous les aspects topologiques et typologiques. Nous proposons également des stratégies pour faire face à l’abandon ou à l’échec de la communication. Des validations basées sur des simulations étendues et des bancs d’essai sont présentées. / In the aerial robotic community, a growing interest for Multi-Robot Systems (MRS) appeared in the last years. This is thanks to i) the technological advances, such as better onboard processing capabilities and higher communication performances, and ii) the promising results of MRS deployment, such as increased area coverage in minimum time. The development of highly efficient and affordable fleet of Unmanned Aerial Vehicles (UAVs) and Micro Aerial Vehicles (MAVs) of small size has paved the way to new large-scale applications, that demand such System of Systems (SoS) features in areas like security, disaster surveillance, inundation monitoring, search and rescue, infrastructure inspection, and so on. Such applications require the robots to identify their environment and localize themselves. These fundamental tasks can be ensured by the exploration mission. In this context, this thesis addresses the cooperative exploration of an unknown environment sensed by a team of UAVs with embedded vision. We propose a multi-robot framework where the key problem is to cooperatively choose specific regions of the environment to be simultaneously explored and mapped by each robot in an optimized manner in order to reduce exploration time and, consequently, energy consumption. Each UAV is able to performSimultaneous Localization And Mapping (SLAM) with a visual sensor as the main input sensor. To explore the unknown regions, the targets – selected from the computed frontier points lying between free and unknown areas – are assigned to robots by considering a trade-off between fast exploration and getting detailed grid maps. For the sake of decision making, UAVs usually exchange a copy of their local map; however, the novelty in this work is to exchange map frontier points instead, which allow to save communication bandwidth. One of the most challenging points in MRS is the inter-robot communication. We study this part in both topological and typological aspects. We also propose some strategies to cope with communication drop-out or failure. Validations based on extensive simulations and testbeds are presented.

Game theory can work as a coordination mechanism in multi-agent robotic systems by representing each robot as a player in a game. In ideal scenarios, game theory algorithms guarantee convergence to optimal configurations and have been widely studied for many applications. However, most of the studies focus on theoretical analysis and lack the details of complete demonstrations. In this regard, we implemented a real-time multi-robot system in order to investigate how game-theoretic methods perform in non-idealized settings. An aerial coverage problem was modeled as a potential game, where each aerial vehicle is an independent decision-making player. These players take actions under limited communication, and each is equipped with onboard vision capabilities. Three game-theoretic methods have been modified and implemented to solve this problem. All computations are performed using onboard devices, independent of any ground entity. The performance of the system is analyzed and compared with different tests and configurations

Robots are no longer constrained to cages in factories and are increasingly taking on roles alongside humans. Before robots can accomplish their tasks in these dynamic environments, they must be able to navigate while avoiding collisions with pedestrians or other robots. Humans are able to move through crowds by anticipating the movements of other pedestrians and how their actions will influence others; developing a method for predicting pedestrian trajectories is a critical component of a robust robot navigation system. A current state-of-the-art approach for predicting pedestrian trajectories is Social-LSTM, which is a recurrent neural network that incorporates information about neighboring pedestrians to learn how people move cooperatively around each other. This thesis extends and modifies that model to output parameters for a multimodal distribution, which better captures the uncertainty inherent in pedestrian movements. Additionally, four novel architectures for representing neighboring pedestrians are proposed; these models are more general than current trajectory prediction systems and have fewer hyper-parameters. In both simulations and real-world datasets, the multimodal extension significantly increases the accuracy of trajectory prediction. One of the new neighbor representation architectures achieves state-of-the-art results while reducing the number of both parameters and hyper-parameters compared to existing solutions. Two techniques for incorporating the trajectory predictions into a planning system are also developed and evaluated on a real-world dataset. Both techniques plan routes that include fewer near-collisions than algorithms that do not use trajectory predictions. Finally, a Python library for Agent-Based-Modeling and crowd simulation is presented to aid in future research.