The state-of-the-art teleimpedance command interfaces used to command the robot stiffness configuration are either too complex to set up, such as those that use physiological signals and other tracking methods or cannot configure the stiffness appropriately for 3d environments. To mitigate these issues, a novel teleimpedance interface is proposed. The proposed interface can independently control the stiffness configuration’s shape, orientation, and size with single-hand operations while allowing the operator to use that hand to command the robot’s position. The teleimpedance interface is attached to the operator’s hand and uses two scroll wheels, a joystick, and a force sensor to configure the robot’s stiffness and has two different modes of operation. Compared to the state-of-the-art methods, the main advantage of the proposed teleimpedance command interface is that it does not require additional hardware with force feedback or complex setup calibrations while allowing for control of the robot’s 3D stiffness configuration with single-handed operation. An experiment with human subjects was performed to demonstrate the proposed interface’s acceptance and functionality. To demonstrate the teleimpedance command interface’s ability to adjust 3D stiffness configurations a teleoperation was performed, utilizing a Kuka robotic arm and a Force Dimension Sigma7 position input interface. The teleimpedance interface functioned as intended during teleoperation in a 3D environment to configure and adjust the 3D stiffness configuration for the task in real-time. The results from the human subject trials indicate that the participants can successfully operate the interface to complete the alignment tasks in both modes for 3D stiffness configurations.

Reinforcement Learning (RL) shows great potential for robotic manipulation tasks, yet it suffers from low sample efficiency and needs extensive exploration of state-action spaces. Some recent methods leverage the commonsense knowledge and reasoning abilities of Large Language Models (LLMs) to guide RL exploration toward more meaningful states. However, LLMs may generate semantically correct but physically infeasible plans, leading to unreliable solutions. In this paper, we propose \textit{Language-Guided exploration for Reinforcement Learning} (LGRL), a novel framework that utilizes LLMs' planning capability to directly guide RL exploration. This approach utilizes LLM planning at both the task and affordance levels, enhancing learning efficiency by directing RL agents toward semantically meaningful actions. Unlike previous methods that rely on the optimality of LLM-generated plans or rewards, LGRL corrects sub-optimality and explores multimodal affordance-level plans without human intervention.
We evaluated LGRL on pick-and-place tasks within standard RL benchmark environments, demonstrating significant improvements in both sample efficiency and success rates.

Behaviour trees (BTs) serve as a powerful hierarchical structure for task execution, simplifying complex tasks but posing challenges in their manual design. The automatic generation of BTs addresses this concern, yet often lacks robust failure recovery options. This study presents Failure Recovery with Ontologically Generated Behaviour Trees (FROGBT), a novel approach bridging this gap by integrating ontological reasoning into the process of automatically generating BTs. This integration establishes a profound link between an agent’s knowledge and its capabilities, offering contextual insights into the agent’s skills. FROGBT enhances skill representation for planning and recovery. The approach’s effectiveness is indicated by its efficiency compared to the state-of-the-art framework for skill-based control, SkiROS, in a similar task. It showcases generality, uniting diverse skills, developed by various engineers, for recurring tasks, and introduces innovative failure recovery strategies. FROGBT highlights ontological reasoning’s potential to enhance BT generation with context-awareness and reasoning abilities, paving the way for future research on failure recovery concepts in generated BTs.

Recent research has shown that a Learning from Demonstration (LfD) approach is useful for teaching robots flexible skills efficiently, and it opens the possibility for non-expert users to program these skills. When learning from demonstration data, learning frameworks should learn representations that are flexible and can generalize to unseen situations. Within the context of multi-reference frame skill learning, this work proposes a framework to learn such a representation without using task-specific heuristics or pre-segmentation of the demonstrations. Local policies are first learned by fitting the local dynamics with respect to each frame using Gaussian Processes (GP). A classifier that determines the relevance of each frame for every time step is then trained in a self-supervised manner. The uncertainty quantification capability of Gaussian Processes is exploited to improve the performance of the local policies and the self-supervised learning process of the classifier. The framework is validated through multi-frame tasks in simulation as well as on a robotic manipulator with a pick-and-place re-shelving task. Its performance is also compared to that of the Task-Parameterised Gaussian Mixture Model (TPGMM) with simulated and robotic data. In this comparison, the proposed model performs better according to metrics that quantify deviation from the goal at each reference frame and according to similarity measures between demonstrations and their corresponding reproductions.

This thesis proposes the novel Behaviour Tree Update Framework (BTUF) for the initial construction and continuous incremental adaptation of Behaviour Trees (BTs) for applications in Learning from Demonstration (LfD) frameworks to create complex robot behaviours associated with Activities of Daily Living (ADL) without requiring the user to have a programming or engineering background. BTUF implements several methods towards that end. Automatic generation of the fundamental structure of BTs within BTUF allows for easy human operation of the framework's Text-based User Interface (TUI). Saving and loading of constructed trees facilitates easy expansion and reusability of constructed trees. By expanding upon an initial base behaviour, seemingly simple behaviours can be adapted to facilitate novel instances thereupon, increasing the complexity and functionality of the constructed tree over time. Experimental validation in the form of a user study has provided proof-of-concept within simulation and has given insight in the initial overall performance and general system acceptance of BTUF. Future work is recommended for further validation and improvement of the proposed framework to one day realise a real-life application within healthcare robotics.

Robotic teleoperation has gained popularity in various applications, including the nuclear industry, where the experience and intelligence of a human operator is necessary for making complex decisions that are beyond the autonomy of robots. Human-robot interfaces that help strengthen an operator’s situational awareness without inducing excessive cognitive load are crucial to the success of teleoperation. This paper presents an intuitive visual interface that allows operators to simultaneously control a 6DOF camera platform and a robotic manipulator whilst experiencing the remote environment through a virtual reality head-mounted-display (VR-HMD). The proposed system, Orbital Head-Mounted Display (OHMD), utilizes head rotation tracking to command camera movement in azimuth and elevation directions around a fixation point located at a robot's end-effector. A human factor study was conducted to compare the interface acceptance, perceived workload, and task performance of OHMD with a conventional interface utilizing multiple fixed cameras (Array) and a standard head-mounted display implementation (HMD). Results show that both the OHMD and HMD interfaces significantly improve task performance, reduce perceived workload and increase interface acceptance compared to the Array interface. Whilst orbital head-mounted display excelled in usefulness, the standard HMD interface allowed operators to perform robotic welding tasks significantly faster. Still participants reported they preferred OHMD due to the increased assistance and freedom in viewpoint selection.

Learning from Demonstration (LfD) aims to learn versatile skills from human demonstrations. The field has been gaining popularity since it facilitates transferring knowledge to robots without requiring much expert knowledge. During task executions, the robot motion is usually influenced by constraints imposed by environments. In light of this, task-parameterized (TP) learning encodes relevant contextual information in reference frames, enabling better skill generalization to new situations. However, most TP learning algorithms require multiple demonstrations in various environment conditions to ensure sufficient statistics for a meaningful model. It is not a trivial task for robot users to create different situations and perform demonstrations under all of them. Therefore, this paper presents a novel concept to learn motion policy from few demonstrations through explicitly solving reference frame weights along the task trajectory. Experimental results in both simulation and real robotic environments validate our approach.

Rehabilitation robotics is a rapidly growing field in the engineering industry. Due to the high repeatability of motion, high therapy costs, and lack of proper quantitative assessment of patient status and progress, rehabilitation centers could benefit from the introduction of robotics. One such device already commercially available is the bodyweight supported treadmill device Lokomat (Hocoma, Switzerland). In this thesis, the kinematics analysis for a modified Lokomat orthosis has been made and validated. These kinematics calculate the Euler angles of the orthosis hip joint as a function of eight degrees of freedom, two degrees of freedom of actuators driving the thigh, and six degrees of freedom of the orthosis pelvis body. The kinematics were validated by using a low-cost HTC Vive VR tracker system, to capture the actual angle of the orthosis hip joint. The kinematics were tested in three experiments. Firstly, the kinematics were tested by moving the orthosis around by hand, tracing the range of motion of the linear actuators. Afterward, the kinematics were validated with a person wearing the device. The first dataset has the person suspended in the air, simulating walking, whilst the final dataset has the person dragging their feet over the ground, to simulate walking. The results from the first experiment indicate that the kinematics calculation tracks the measured angles with an RMSE of less than 6% of the total range of motion. Later experiments suffered from drift in the Vive trackers, mainly in the flexion angles, worsening performance. Compensating for this drift shows the kinematics to still be similarly accurate to the first experiment.

Inertial Measurement Unit (IMU)-based motion capture has gained interest over the years due to its potential to measure human movement in the clinic and on the sports field at low cost. Still, IMU-based motion reconstruction remains a challenging task as these IMU measurements are corrupted by noise and bias. There have been many filtering and sensor fusion algorithms developed to address this problem, but limited attention has been paid to including the system dynamics of the subject to estimate kinematics and kinetics from multiple IMUs. I propose a novel motion reconstruction algorithm for capturing 3D motions in a markerless and unconstrained environment using gyroscope and accelerometer measurements. The algorithm is based upon an Iterated Extended Kalman Filter for state estimation and the 3D dynamical model of the subject created in OpenSim (IEKF-OS). The IEKF-OS algorithm consists of two stages. The first stage incorporates the dynamics of the subject to predict the motion. The second stage tracks measurements from multiple IMUs to improve model estimates of joint angles and speeds. The goal of this thesis is threefold. Firstly, the IEKF-OS is derived and verified using simulations performed on a six-link robot manipulator including sensor placement errors. Secondly, experiments are performed to validate the algorithm’s estimations against the robot’s ground truth joint encoder values. Thirdly, the algorithm is compared against an inverse kinematics method to track IMU estimated orientations (OpenSense) provided by OpenSim. The IEKF-OS algorithm showed lower motion tracking errors compared to OpenSense for various motions performed by the robot manipulator. The joint angle estimations as computed by both methods are compared against the gold-standard ground truth robot encoder values. OpenSense joint angle values were in the range of 0.6-6.4 [deg] RMSE, whereas the IEKF-OS algorithm estimated joint angles and speeds in the range of 0.4-2.8 [deg] and 0.3-2 [deg/s] RMSE, respectively. These results highlight accurate 3D motion reconstruction on a six-link robot manipulator. Contrary to OpenSense, the IEKF-OS is able to accurately reconstruct motions regardless of magnetic disturbances because it does not rely on magnetometer data.

This study proposes the creation of a multi-modal feedback system to guide humans towards ergonomic poses. A number of studies have tried to come up with methods where subjects are alerted upon crossing biomechanical or ergonomic thresholds while doing a task but not many have tried to successfully and efficiently guide users to ergonomic positions after having alerted them. Through this study we propose the creation of a multi-modal feedback system comprising of a visual and a speech based audio feedback and hypothesize that the proposed system will lead to a better performance as compared to the other feedback modalities when trying to guide users from one pose to another. During our study we have conducted two sets of experiments to carry out a comparative study between only audio, only visual and the proposed multi-modal feedback system to try and find the modality most effective and successful in guiding humans for pose corrections and a comparative study between two types of speech based audio feedbacks in joint space and end point space to motivate our choice for using the more desired one between the two for our proposed system.
Speech based feedback in joint space came out as the preferred audio feedback due to its ability to allow users to carry out efficient and coordinated inter-joint movements especially in cases of high redundancy whereas the proposed multi-modal feedback system successfully shows its superiority over the other feedback
modalities by showing equivalent results against the benchmark visual feedback when measured objectively and better results when measured subjectively due to its ability to successfully combine the advantages of audio and visual feedback and at the same time, avoid their limitations.

Imitation Learning is a technique that enables programming the behavior of agents through demonstration, as opposed to manually engineering behavior. However, Imitation Learning methods require demonstration data (in the form of state-action labels) and in many scenarios, the demonstrations can be expensive to obtain or too complex for a demonstrator to execute. This lack or sub-optimality of demonstrations limits the applicability and performance of many Imitation Learning methods.

Advancements in Interactive Imitation Learning techniques however, have made it easier for demonstrators to train agents and improve their performance. These techniques involve demonstrators interacting with and guiding the agent as it performs the requisite task. This guidance is typically in the form of corrections or feedback on the current actions being executed by the agent.

In this thesis, a novel Interactive Learning technique is proposed that uses human corrective feedback in state-space to train and improve agent behavior. This technique is beneficial since providing guidance to the agent in terms of `changing its state' is often easier or more intuitive for the human demonstrator (as opposed to changing the actions being executed). For instance, in manipulation tasks using a robotic arm, it is easier for the demonstrator to provide state information such as the Cartesian position of the end-effector rather than low-level action information such as joint angles. Keeping such scenarios in mind, we propose our method titled: Teaching Imitative Policies in State-space (TIPS).

We evaluate the performance of TIPS for various control tasks as part of the OpenAI Gym toolkit as well as for a manipulation task using a KUKA LBR iiwa robotic arm. We show that through continuous improvement via feedback, agents trained using TIPS outperform the demonstrator and in-turn outperform conventional Imitation Learning agents.

Introduction: Pes equinus with increased ankle joint stiffness is a common impairment in stroke and Cerebral Palsy (CP) patients with structural and/or neural deficits of the ankle muscles. Equinus gait is characterized by toe-strike, abnormal ankle plantarflexion and decreased ankle range of motion (ROM). A new ankle-foot orthosis (AFO) with negative stiffness was previously developed to compensate for the increased ankle joint stiffness and improve equinus gait, reducing plantarflexion and increasing ankle ROM. The goal of this project was to predict the effects of ankle stiffness compensation on equinus gait simulations.

Methods: Forward simulations of unimpaired and equinus gait were generated in
the sagittal plane with a musculoskeletal model implemented in OpenSim and a
neural controller implemented in SCONE. After validation of the unimpaired gait
simulation, a sensitivity analysis on ankle kinematics was performed, introducing
structural and/or neural alterations of the ankle muscles to achieve an equinus gait. Consecutively, equinus gait simulations were validated with previously collected data of CP patients. Finally, an AFO model was developed in OpenSim to simulate AFO-assisted gait.

Results: The unimpaired gait simulation yielded realistic results, and was robust to all alterations, generating stable gaits. Shorter Gastrocnemius fiber length and/or increased plantarflexor muscles activations, with Tibialis Anterior weakness, had the largest effect on the ankle joint kinematics and resulted in realistic CP equinus gaits. Simulations of AFO-assisted gait resulted in reduced abnormal plantarflexion, and increased ankle ROM in the condition with shorter Gastrocnemius.

Discussion and Conclusion: This study presented realistic simulations of equinus
gait, by modelling structural and/or neural alterations of the ankle muscles and
predicted improved ankle function when compensating ankle stiffness with an external force. Although higher stiffness compensation and more simulations should be obtained, this study provides a solid ground for further investigation of the AFO effects. These results can ultimately assist in the prescription and tuning of the new AFO for patients with equinus.

Monitoring selectively directed auditory attention in groups can be helpful in a range of contexts, such as in education. In real-world settings, selectively directed attention cannot be monitored by relating physiological signals to known event markers. Determining the similarity of physiological responses across individuals in a group (i.e., physiological synchrony - PS) may provide a solution to this problem, as it has been shown reflective of shared attentional engagement toward audiovisual stimuli. This study was aimed at examining whether PS in autonomic and neural measures are suitable markers to recover selectively directed attentional focus and moments of emotional or task-related relevance in a composite auditory stimulus.
Electroencephalography (EEG), electrodermal activity (EDA) and cardiac inter-beat interval (IBI) were monitored from participants who heard a composite auditory stimulus, consisting of a narrative
audiobook, interspersed with short stimuli. One group of the participants (n = 13) was instructed to attend to the narrative, and the other group (n = 13) was instructed to attend to the short stimuli,
that were tones that attending participants needed to keep track of, emotional sounds and a stress test attending participants needed to perform.
EEG and EDA signals of participants were more strongly synchronized with those of participants in the same attentional condition than with those of participants in the other attentional condition. No such effect was found in IBI. For a single individual, PS in EEG allowed attribution to the correct attentional group in 85% of the cases, for EDA this was 81% and for IBI accuracy was below chance level. Further analyzing EEG, PS was higher during relevant stimulus presentation than over the entire audiobook. PS across stimulus-attending participants was higher than across book-attending participants during these short stimuli.
It is concluded that PS in both autonomic and neural measures can be suitable to recover selectively directed attention, as indicated by the high classification accuracy for EEG and EDA. The results of
this study also indicate that PS may be suited for the detection of moments of emotional or task-related relevance, as PS in EEG was higher during relevant stimulus presentation than over the entire experiment. Future research should investigate whether PS in autonomic measures could also be suited for detection of relevant events.

Tensegrity is a structural form that is defined as a set of rigid elements suspended in a net of continuous tension. This structure shows potential for compliance, impact tolerance and mechanical robustness. However, its non-linear coupled dynamics and often complex geometry require advanced control strategies. An actuator reference planning strategy to bring tensegrity robots closer to controlled full body movements was proposed by Guido Tournois \cite{GuidoThesis} in 2017. This strategy, called the Full Body Reference Planner (FBRP), finds a sequence of equilibrium configurations for a tensegrity structure, predominantly to follow a trajectory in space. However, the method is incapable of incorporating inequality constraints while obtaining said equilibrium configurations. This is a problem when dealing with certain restrictions, e.g., actuator limitations and stability of the structure.

In this thesis we implemented a robust way to account for inequality constraints while utilizing the FBRP. That was done by means of optimization, i.e., an implementation of a Sequential Quadratic Programming method to ensure inequality constraints were respected for each configuration. The approach was validated in scenarios related to practical applications where inequality constraints were enforced. The results showed advancements towards practical feasibility. Furthermore, the robustness, efficiency and accuracy of the method were validated. The extended implementation depicted robustness to parameter variations and good results in terms of accuracy. However, given the iterative nature of the method, it was more computationally expensive than its precursor.