Master of Science, The Ohio State University, 2013, Electrical and Computer Engineering

When robots are integrated into the real world, chances are they will not be able to completely avoid situations in which they are bumped or pushed unexpectedly. In these situations, the robot could potentially damage itself, damage its surroundings, or fail to perform its tasking unless it is able to take active countermeasures to prevent or recover from falling. One such countermeasure, referred to as reactive stepping, involves a robot taking a series of steps in order to regain balance and recover from a push. Research into reactive stepping typically focuses on choosing which step to take. This thesis proposes a machine learning approach to reactive stepping. This approach leverages neural networks to calculate a series of steps that return the robot to a stable position. It was theorized that the robot would become stable if it always chose the step resulting in the highest reduction of energy. Theories were tested using a compass model that incorporated parameters and constraints realistic of an actual humanoid robot. The machine learning approach using neural networks performed favorably in both computation time and push recovery effectiveness when compared with the linear least squares, nearest interpolation, and linear interpolation methods. Results showed that when using neural networks to calculate the best step for an arbitrary push within the defined range, the compass model was able to successfully recover from 97% of the pushes applied. The procedure was kept very general and could be used to implement reactive stepping on physical robots, or other robot models.

Committee: Yuan Zheng (Advisor); David Orin (Committee Member) Subjects: Computer Science; Engineering; Robotics

Master of Science, The Ohio State University, 2021, Mechanical Engineering

Bipedal robots are becoming more popular in performing tasks in an environment that is designed for humans. For this purpose, most bipedal robots are equipped with various sensors to sense the robot's environment. From the measurements of the sensors, a perception system is implemented to translate and convert the raw data into a meaningful format corresponding to the tasks and also provide safety for humans, properties in the environment as well as the robot itself. This thesis presents the implementation of a perception system using various sensors available to a bipedal robot, Digit, to obtain objectively useful information of the environment as well as the state of the robot itself. Various methods of data processing were applied to available sensor measurements, then a mapping algorithm was implemented to generate a 3D model of the environment. Simultaneous localization and mapping (SLAM) algorithm was also implemented to perform mapping and provide odometry for localization in the absence of an external source of odometry. We found that performing SLAM using Light Detection and Ranging sensor (LiDAR) performs exceptionally well on the bipedal robot in closed indoor space. Additionally, state estimation is implemented with Invariant Extended Kalman filter using inertial measurement data and the assumption of contact points to predict the state of the robot over time. The performance of position estimation from Invariant Extended Kalman filter and odometry from LiDAR SLAM is compared with the default state estimator from Digit itself which are demonstrated through an experiment with ground truth reference.

Committee: Keith Redmill (Committee Member); Ayonga Hereid (Advisor) Subjects: Mechanical Engineering; Robotics

Master of Science in Electrical Engineering (MSEE), Wright State University, 2019, Electrical Engineering

A legged robot with three-segmented limbs is used to study the effects of leg compliance originating from the joint level on the stability of hopping in place and running. The three-segments allow each leg to be kinematically configured an infinite number ways that satisfy the desired landing condition parameters, total leg length and angle. These two parameters along with the amount of energy thrust during stance determine the motion of the system during a single stride. The goal of this work is to explore the potential values for the leg parameters of three-segment leg, that provide additional stability when compared to legs with fewer segments. The stability is analyzed based on the how well the robot can return to the desired height while hopping and the desired velocity while running. Given a fixed point in the control space, where the system returns to the initial height and velocity, the stability of different leg configurations is compared by counting the number of steps the robot can take before falling over. The added thrust to the joints and the leg attack angles are varied to observe the stability regions for different kinematic configurations, and compared to biped with lower number of leg segments, to prove how leg segmentation provides additional stability. It may also be useful to perform the same research for running on uneven terrains.

Committee: Luther Palmer III Ph.D. (Advisor); Xiaodong Zhang Ph.D. (Committee Member); Arnab K. Shaw Ph.D. (Committee Member) Subjects: Electrical Engineering; Robotics