CPG-Based oscillator networks are increasingly being used to drive multi-limbed robots. To produce effective gaits with these networks, the relationship between the CPG parameters and the characteristics of the gait must be determined. However, due to the nonlinear nature of the oscillators, this relationship is challenging to ascertain. In this work a reinforcement learning algorithm is used to determine the CPG parameters that produce propulsively beneficial kinematics in a multi-fin underwater robot. Due to the high computational cost in creating high fidelity simulations of underwater systems, an alternate method using a low fidelity simulation is explored. To better simulate the dynamics of a two-finned swimming robot a thorough force sweep is conducted on the subject robot in a controlled environment. The resulting force data is used as the dynamic information in a simple simulation. This method allows for the learning of CPG weight settings that produce desired kinematic operating conditions and their resulting forces in simulation. Using this method, when the learned CPG parameters were applied directly to the physical robot, the robot executed the same desired kinematics and forces as expected from simulation with no additional learning needed.

Fish are capable of producing a wide repertoire of 3D propulsive forces using their fins, and have inspired the development of compliant, multiple-DOF, robotic fins with similar capabilities. Most of these robotic fins are under open-loop control on propulsive force because the forces are challenging to model. Understanding how to predict propulsive forces for these types of fins would significantly advance the state of the art towards closed-loop control of forces. Distributed sensors within robotic fins have been used to predict propulsive forces using linear models, but these models fail to predict forces when fin kinematics become more complex. The objective of the work presented herein is to understand the use of nonlinear, multiple-input-single-output (MISO) Volterra series models between intrinsic sensory measurements and propulsive forces of a flexible robotic fin. Techniques in nonlinear system identification are used to address model conditioning. Nonlinear models predict the propulsive forces well, capturing features of both thrust and lateral forces. Nonlinear models significantly outperformed linear models both in cost of implementation and performance. The best sensor sampling practice was to sample from multiple locations with both pressure and bending modalities. Distributed sensing paired with nonlinear Volterra series models was successful for predicting the forces created by flexible robotic fins with complex kinematics and multiple degrees of freedom.

Biologically-inspired robotic (biorobotic) platforms have been successfully adapted for engineering use, but it is difficult to extend these platforms' locomotive gaits to meet optimization goals. The gait spaces of biorobotic platforms can be very large, with multiple local optima and intractable numerical models. Further, the time cost of empirical exploration is often prohibitive. Micro-genetic algorithms have been successful in developing inverse kinematics in simulation, optimizing in spaces with numerous local optima, and working quickly to optimize with low numbers of trials, but have not yet been evaluated for online robotic gait development. To address the problem of engineering gait development in a biorobotic space, a micro-genetic algorithm (μGA) is evaluated on a biorobotic pectoral fin platform. The μGA effectively optimizes in the gait space with low time costs, discovering new gaits that optimize thrust force production on the swimming fin. The μGA also reveals parameter tuning strategies for changing propulsive forces. Overall, the μGA framework is shown to be effective at online optimization in a large, complex biorobotic gait space.

A comprehensive understanding of the ways in which fish create and control forces is fundamental to engineering underwater vehicles that maneuver with the agility of fish. In this study the sunfish was selected as a biological model from which to understand pectoral fin motions and forces during hover. The kinematic patterns of the biological fin were identified and implemented on a biorobotic model of the fin. The effects of fin patterns and mechanical properties on force were evaluated. Pressure was measured at multiple points on the fin's surface and assessed for use in the closed loop control of fin force. The study revealed that a wide range of motions are used during hover, and that forces are significantly different from those found previously for steady swimming. However as fin speeds increase, the fin's dynamic motions, and the magnitude and direction of the forces become more similar to those of steady swimming. Collective measures of pressure over the fin's surface exhibited trends that correlated well with fin forces in relative magnitudes and directions. Results strongly suggest that distributed measures of pressure are useful for force prediction and control.

The remarkable swimming abilities of bony fish are the result of multiple interacting subsystems, each tuned to perform certain roles. These subsystems, which include the statically unstable body, multiple highly actuated fins, oscillatory neural controllers, and distributed senses, are not often studied as mutually dependent systems. This research program is developing biorobotic models of these systems and integrating the systems into a biorobotic fish so that interdependencies can be explored during free swimming. The robot body was derived from a bluegill sunfish, and has a tunable mass distribution and a mixture of rigid and flexible sections so that dynamical characteristics of the fish body can be explored. Five highly deformable fins have structural properties scaled to those of biological fins and can create gait patterns for steady swimming and maneuvers. A first generation artificial CPG has been programmed for each fin on a network of five low power microcontrollers. Finally, a dedicated biorobotic pectoral fin has been developed and instrumented with distributed sensory systems so relevant physical (e.g., fin curvature) and hydrodynamic (e.g., pressure) data can be identified and used to predict fin force for closed loop control.

Experimental analyses of propulsion in freely-swimming fishes have led to the development of self-propelling pectoral and caudal fin robotic devices. These biorobotic models have been used in conjunction with biological and numerical studies to investigate the effects of the fin's kinematic patterns and structural properties on forces and flows. Data from both biorobotic fins was presented and discussed in terms of the utility of using robotic models for understanding fish locomotor dynamics. Through the use of the robotic fins, it was shown that subtle changes to the kinematics and/or the mechanical properties of fin rays can impact significantly the magnitude, direction, and time course of the 3D forces used in propulsion and maneuvers.

A series of biorobotic fins has been developed based on the pectoral fin of the bluegill sunfish. These robotic fins model physical properties of the biological fin, and execute kinematics derived from sunfish motions that were identified to be most responsible for thrust. When the physical properties of the robotic fin are tuned appropriately to operating conditions, the robotic fin, like the sunfish, produces positive thrust throughout the entire fin beat. Due to having many degrees of freedom, these fins can be used to generate and control forces for propulsion and maneuvering.