Masters Degrees (Electrical and Electronic Engineering)

Permanent URI for this collection

https://scholar.sun.ac.za/handle/10019.1/766

Browse

Browsing Masters Degrees (Electrical and Electronic Engineering) by Subject "3D turning analysis"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    3D turning analysis of a Bipedal Robot
    (Stellenbosch : Stellenbosch University, 2022-04) Pretorius, Dean; Fisher, Callen; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT: There is stark contrast between the abilities of legged locomotion found in nature, and locomotion found in lab environments. This performance gap is indicative of a large knowledge gap. Roboticists are required to bridge these gaps to truly invite robots to detach from their support rigs, and actuate within the real world. In this thesis, non-planar contact and discontinuous locomotive dynamics were modeled as a trajectory optimization problem. Consequently, this made understanding the complexities of legged locomotion more tractable. Understanding, and being able to leverage, contact is crucial to successful legged locomotion. Therefore, a comprehensive investigation was conducted into non-planar contact dynamics using a monopod robot. Here, methods of modeling the Coulomb friction cone in contact implicit trajectory optimization were implemented. Literature suggests replacing the friction cone with a polyhedral approximation thereof. However, this method is known to underestimate the resultant friction in non-planar environments. This thesis presents a novel method of modeling the 3D friction cone and compares it to an implementation of the polyhedral approximation. Results from this comparison show that the novel method was significantly more computationally efficient than the polyhedral approximation, without underestimating the friction cone. Dynamic bipedal locomotion remains a struggle for most robotic platforms. Robotics literature provides few examples of robots achieving agile, dynamic locomotion. Therefore, trajectories realizing non-planar dynamic bipedal motion were generated. Experiments were conducted into acceleration, steady-state, deceleration, and rapid turning off the sagittal plane. Optimal trajectories displayed the robot walking at speeds resulting in a Froude number less than 0.5, and running at speeds resulting in a higher Froude number. This is consistent with dynamic gaits found in nature. A sliding-mass velocity profile emerged when conducting long-time-horizon trajectories where the robot accelerated from a rest position and decelerated back to rest after completing multiple steps in a periodic steady-state gait. Additionally, when turning off the sagittal plane, slip occurred at least 93.32% of the duration of contact, and turn overshoot is present in all turn trajectories.