Soft continuum robots with the ability to bend, twist, elongate, and shorten, similar to octopus arms, have many potential applications, such as dexterous manipulation and navigation through unstructured, dynamic environments. Novel soft materials such as smart hydrogels, which change volume and other properties in response to stimuli such as temperature, pH, and chemicals, can potentially be used to construct soft robots that achieve self-regulated adaptive reconfiguration through on-demand dynamic control of local properties. However, the design of controllers for soft continuum robots is challenging due to their high-dimensional configuration space and the complexity of modeling soft actuator dynamics. To address these challenges, this dissertation presents two different model-based control approaches for robots with distributed soft actuators and sensors and validates the approaches in simulations and physical experiments. It is demonstrated that by choosing an appropriate dynamical model and designing a decentralized controller based on this model, such robots can be controlled to achieve diverse types of complex configurations. The first approach consists of approximating the dynamics of the system, including its actuators, as a linear state-space model in order to apply optimal robust control techniques such as H∞ state-feedback and H∞ output-feedback methods. These techniques are designed to utilize the decentralized control structure of the robot and its distributed sensing and actuation to achieve vibration control and trajectory tracking. The approach is validated in simulation on an Euler-Bernoulli dynamic model of a hydrogel based cantilevered robotic arm and in experiments with a hydrogel-actuated miniature 2-DOF manipulator. The second approach is developed for soft continuum robots with dynamics that can be modeled using Cosserat rod theory. An inverse dynamics control approach is implemented on the Cosserat model of the robot for tracking configurations that include bending, torsion, shear, and extension deformations. The decentralized controller structure facilitates its implementation on robot arms composed of independently-controllable segments that have local sensing and actuation. This approach is validated on simulated 3D robot arms and on an actual silicone robot arm with distributed pneumatic actuation, for which the inverse dynamics problem is solved in simulation and the computed control outputs are applied to the robot in real-time.

Mechanical impedance is a concept that is used to model biomechanical propertiesof human joints. These models can then be utilized to provide insight into the inner workings of the human neuromuscular system or to provide insight into how to best design controllers for robotic applications that either attempt to mimic capabilities of the human neuromuscular system or physically interact with it. To further elucidate patterns and properties of how the human neuromuscular system modulates mechanical impedance at the human ankle joint, multiple studies were conducted. The first study was to assess the ability of linear regression models to characterize the change in stiffness - a component of mechanical impedance - seen at the human ankle during the stance phase of walking in the Dorsiflexion-Plantarflexion (DP) direction. A collection of biomechanical variables were used as input variables. The R^2 value of the best performing model was 0.71. The second and third studies were performed to showcase the ability of a newly developed twin dual-axis platform, which goes beyond the limits of a single dual-axis platform, to quantify bilateral stiffness properties. The second study quantified the bilateral mechanical stiffness of the human ankle joint for healthy able-bodied subjects during the stance phase of walking and during quiet standing in both the DP and inversion-eversion directions. Subjects showed a high level of subject specific symmetry. Lastly, a similar bilateral ankle characterization study was conducted on a set of subjects with multiple sclerosis, but only during quiet standing and in the DP direction. Results showed a high level of discrepancy between the subject’s most-affected and least-affected limbs with a larger range and variance than in the healthy population.

Technological progress in robot sensing, design, and fabrication, and the availability of open source software frameworks such as the Robot Operating System (ROS), are advancing the applications of swarm robotics from toy problems to real-world tasks such as surveillance, precision agriculture, search-and-rescue, and infrastructure inspection. These applications will require the development of robot controllers and system architectures that scale well with the number of robots and that are robust to robot errors and failures. To achieve this, one approach is to design decentralized robot control policies that require only local sensing and local, ad-hoc communication. In particular, stochastic control policies can be designed that are agnostic to individual robot identities and do not require a priori information about the environment or sophisticated computation, sensing, navigation, or communication capabilities. This dissertation presents novel swarm control strategies with these properties for detecting and mapping static targets, which represent features of interest, in an unknown, bounded, obstacle-free environment. The robots move on a finite spatial grid according to the time-homogeneous transition probabilities of a Discrete-Time Discrete-State (DTDS) Markov chain model, and they exchange information with other robots within their communication range using a consensus (agreement) protocol. This dissertation extend theoretical guarantees on multi-robot consensus over fixed and time-varying communication networks with known connectivity properties to consensus over the networks that have Markovian switching dynamics and no presumed connectivity. This dissertation develops such swarm consensus strategies for detecting a single feature in the environment, tracking multiple features, and reconstructing a discrete distribution of features modeled as an occupancy grid map. The proposed consensus approaches are validated in numerical simulations and in 3D physics-based simulations of quadrotors in Gazebo. The scalability of the proposed approaches is examined through extensive numerical simulation studies over different swarm populations and environment sizes.

Autonomous Robots have a tremendous potential to assist humans in environmental monitoring tasks. In order to generate meaningful data for humans to analyze, the robots need to collect accurate data and develop reliable representation of the environment. This is achieved by employing scalable and robust navigation and mapping algorithms that facilitate acquiring and understanding data collected from the array of on-board sensors. To this end, this thesis presents navigation and mapping algorithms for autonomous robots that can enable robot navigation in complexenvironments and develop real time semantic map of the environment respectively. The first part of the thesis presents a novel navigation algorithm for an autonomous underwater vehicle that can maintain a fixed distance from the coral terrain while following a human diver. Following a human diver ensures that the robot would visit all important sites in the coral reef while maintaining a constant distance from the terrain reduces heterscedasticity in the measurements. This algorithm was tested on three different synthetic terrains including a real model of a coral reef in Hawaii. The second part of the thesis presents a dense semantic surfel mapping technique based on top of a popular surfel mapping algorithm that can generate meaningful maps in real time. A semantic mask from a depth aligned RGB-D camera was used to assign labels to the surfels which were then probabilistically updated with multiple measurements. The mapping algorithm was tested with simulated data from an RGB-D camera and the results were analyzed.

Many medical procedures, like surgeries, deal with the physical manipulation of sensitive internal tissues. Over time, new medical tools and techniques have been developed to improve the safety and efficacy of these procedures. Despite the leaps and bounds of progress made up to the present day, three major obstacles (among others) persist, bleeding, pain, and the risk of infection. Advances in minimally invasive treatments have transformed many formerly risky surgical procedures into very safe and highly successful routines. Minimally invasive surgeries are characterized by small incision profiles compared to the large incisions in open surgeries, minimizing the aforementioned issues. Minimally invasive procedures lead to several benefits, such as shorter recovery time, fewer complications, and less postoperative pain. In minimally invasive surgery, doctors use various techniques to operate with less damage to the body than open surgery. Today, these procedures have an established, successful history and promising future. Steerable needles are one of the tools proposed for minimally invasive operations. Needle steering is a method for guiding a long, flexible needle through curved paths to reach targets deep in the body, eliminating the need for large incisions. In this dissertation, we present a new needle steering technology: magnetic needle steering. This technology is proposed to address the limitations of conventional needle steering that hindered its clinical applications. Magnetic needle steering eliminates excessive tissue damage, restrictions of the minimum radius of curvature, and the need for a complex nonlinear model, to name a few. It also allows fabricating the needle shaft out of soft and tissue-compliant materials. This is achieved by first developing an electromagnetic coil system capable of producing desired magnetic fields and gradients; then, a magnetically actuated needle is designed, and its effectiveness is experimentally evaluated. Afterward, the scalability of this technique was tested using permanent magnets controlled with a robotic arm. Furthermore, different configurations of permanent magnets and their influence on the magnetic field are investigated, enabling the possibility of designing a desired magnetic field for a specific surgical procedure and operation on a particular organ. Finally, potential future directions towards animal studies and clinical trials are discussed.

Regolith excavation systems are the enabling technology that must be developed in order to implement many of the plans for in-situ resource utilization (ISRU) that have been developed in recent years to aid in creating a lasting human presence on the surface of the Moon, Mars, and other celestial bodies. The majority of proposed ISRU excavation systems are integrated onto a wheeled mobility system, however none yet have proposed the use of a screw-propelled vehicle, which has the potential to augment and enhance the capabilities of the excavation system. As a result, CASPER, a novel screw-propelled excavation rover is developed and analyzed to determine its effectiveness as a ISRU excavation system. The excavation rate, power, velocity, cost of transport, and a new parameter, excavation transport rate, are analyzed for various configurations of the vehicle through mobility and excavation tests performed in silica sand. The optimal configuration yielded a 28.4 kg/hr excavation rate and11.2 m/min traverse rate with an overall system mass of 3.4 kg and power draw of26.3 W. CASPER’s mobility and excavation performance results are compared to four notable proposed ISRU excavation systems of various types. The results indicate that this architecture shows promise as an ISRU excavator because it provides significant excavation capability with low mass and power requirements.

As technological advancements in silicon, sensors, and actuation continue, the development of robotic swarms is shifting from the domain of science fiction to reality. Many swarm applications, such as environmental monitoring, precision agriculture, disaster response, and lunar prospecting, will require controlling numerous robots with limited capabilities and information to redistribute among multiple states, such as spatial locations or tasks. A scalable control approach is to program the robots with stochastic control policies such that the robot population in each state evolves according to a mean-field model, which is independent of the number and identities of the robots. Using this model, the control policies can be designed to stabilize the swarm to the target distribution. To avoid the need to reprogram the robots for different target distributions, the robot control policies can be defined to depend only on the presence of a “leader” agent, whose control policy is designed to guide the swarm to a particular distribution. This dissertation presents a novel deep reinforcement learning (deep RL) approach to designing control policies that redistribute a swarm as quickly as possible over a strongly connected graph, according to a mean-field model in the form of the discrete-time Kolmogorov forward equation. In the leader-based strategies, the leader determines its next action based on its observations of robot populations and shepherds the swarm over the graph by probabilistically repelling nearby robots. The scalability of this approach with the swarm size is demonstrated with leader control policies that are designed using two tabular Temporal-Difference learning algorithms, trained on a discretization of the swarm distribution. To improve the scalability of the approach with robot population and graph size, control policies for both leader-based and leaderless strategies are designed using an actor-critic deep RL method that is trained on the swarm distribution predicted by the mean-field model. In the leaderless strategy, the robots’ control policies depend only on their local measurements of nearby robot populations. The control approaches are validated for different graph and swarm sizes in numerical simulations, 3D robot simulations, and experiments on a multi-robot testbed.

JOOEE is a cube-shaped lunar robot with a simple yet robust design. JOOEE ishermetically sealed from its environment with no external actuators. Instead, JOOEE spins three internal orthogonal flywheels to accumulate angular momentum and uses a solenoid brake at each wheel to transfer the angular momentum to the body. This procedure allows JOOEE to jump and hop along the lunar surface. The sudden transfer in angular momentum during braking causes discontinuities in JOOEE’s dynamics that are best described using a hybrid control framework. Due to the irregular methods of locomotion, the limited resources on the lunar surface, and the unique mission objectives, optimal control profiles are desired to minimize performance metrics such as time, energy, and impact velocity during different maneuvers. This paper details the development of an optimization tool that can handle JOOEE’s dynamics including the design of a hybrid control framework, dynamics modeling and discretization, optimization cost functions and constraints, model validation, and code acceleration techniques.

Chapter 1: Functional Specialization and Arm Length in Octopus bimaculoides<br/>Although studies are limited, there is some evidence that octopuses use their arms for specialized functions. For example, in Octopus maya and O. vulgaris, the anterior arms are utilized more frequently for grasping and exploring (Lee, 1992; Byrne et al., 2006a), while posterior arms are more frequently utilized for crawling in O. vulgaris (Levy et al., 2015). In addition, O. vulgaris uses favored arms when retrieving food and making contact with a T-maze as dictated by their lateralized vision (Byrne, 2006b). O. vulgaris also demonstrates a preference for anterior arms when retrieving food from a Y-maze (Gutnick et. al. 2020). In Octopus bimaculoides bending and elongation were more frequent in anterior arms than posterior arms during reaching and grasping tasks, and right arms displayed deformation more frequently than left arms, with the exception of the hectocotylus (R3) in males (Kennedy et. al. 2020). Given these observed functional differences, the goal of this study was to determine if morphological differences exist between different octopus arm identities, coded as L1-L4 and R1-R4. In particular, the relationship between arm length and arm identity was analyzed statistically. The dataset included 111 intact arms from 22 wild-caught specimens of O. bimaculoides (11 male and 11 female). Simple linear regressions and an analysis of covariance were performed to test the relationship between arm length and a number of factors, including body mass, sex, anterior versus posterior location, and left versus the right side. Mass had a significant linear relationship with arm length and a one-way ANOVA demonstrated that arm identity is significantly correlated with arm length. Moreover, an analysis of covariance demonstrated that independent of mass, arm identity has a significant linear relationship with arm length. Despite an overall appearance of bilateral symmetry, arms of different identities do not have statistically equivalent lengths in O. bimaculoides. Furthermore, differences in arm length do not appear to be related to sex, anterior versus posterior location, or left or right side. These results call into question the existing practice of treating all arms as equivalent by either using a single-arm measurement as representative of all eight or calculating an average length and suggest that morphological analyses of specific arm identities may be more informative.<br/><br/>Chapter 2: Predicting and Analyzing Octopus bimaculoides Sensitivity to Global Anesthetic<br/>Although global anesthetic is widely used in human and veterinary medicine the mechanism and impact of global anesthetic is relatively poorly comprehended, even in well-studied mammalian models. Invertebrate anesthetic is even less understood. In order to evaluate factors that impact anesthetic effectiveness analyses were conducted on 22 wild-caught specimens of Octopus bimaculoides during 72 anesthetic events.Three machine learning models: regression tree, random forest, and generalized additive model were utilized to make predictions of the concentration of anesthetic (percent ethanol by volume) from 11 features and to determine feature importance in making those predictions. The fit of each model was analyzed on three criteria: correlation coefficient, mean squared error, and relative error. Feature importance was determined in a model-specific manner. Predictions from the best performing model, random forest, have a .82 correlation coefficient with experimental values. Feature importance suggests that temperature on arrival and cohabitation factors strongly influence predictions for anesthesia concentration. This likely indicates the transportation process was incurring stress on the animals and that cohabitation was also stressful for the typically solitary O. bimaculoides. This long-term stress could lead to a decline in the animal’s well-being and a lower necessary ethanol concentration (Horvath et al., 2013). This analysis provides information to improve the care of octopus in laboratory settings and furthers the understanding of the effects of global anesthetic in invertebrates, particularly one with a distributed nervous system.

Stroke is a debilitating disorder and 75% of individuals with stroke (iwS) have upper extremity deficits. IwS are prescribed therapies to enhance upper-extremity mobility, but current most effective therapies have minimum requirements that the individuals with severe impairment do not meet. Thus, there is a need to enhance the therapies. Recent studies have shown that StartReact -the involuntary release of a planned movement, triggered by a startling stimulus (e.g., loud sound)- elicits faster and larger muscle activation in iwS compared to voluntary-initiated movement. However, StartReact has been only cursorily studied to date and there are some gaps in the StartReact knowledge. Previous studies have only evaluated StartReact on single-jointed movements in iwS, ignoring more functional tasks. IwS usually have abnormal flexor activity during extension tasks and abnormal muscle synergy especially during multi-jointed tasks; therefore, it is unknown 1) if more complex multi-jointed reach movements are susceptible to StartReact, and 2) if StartReact multi-jointed movements will be enhanced in the same way as single-jointed movements in iwS. In addition, previous studies showed that individuals with severe stroke, especially those with higher spasticity, experienced higher abnormal flexor muscle activation during StartReact trials. However, there is no study evaluating the impact of this elevated abnormal flexor activity on movement, muscle activation and muscle synergy alterations during voluntary-initiated movements after exposure to StartReact.
This dissertation evaluates StartReact and the voluntary trials before and after exposure to StartReact during a point-to-point multi-jointed reach task to three different targets covering a large workspace. The results show that multi-jointed reach tasks are susceptible to StartReact in iwS and the distance, muscle and movement onset speed, and muscle activations percentages and amplitude increase during StartReact trials. In addition, the distance, accuracy, muscle and movement onsets speeds, and muscle synergy similarity indices to the norm synergies increase during the voluntary-initiated trials after exposure to StartReact. Overall, this dissertation shows that exposure to StartReact did not impair voluntary-initiated movement and muscle synergy, but even improved them. Therefore, this study suggests that StartReact is safe for more investigations in training studies and therapy.