THESIS DEFENSE: Self-Assembling Modular Systems: Enhancing Efficiency and Accuracy in Robotic Lattice Assembly

Speaker: John Romanishin , CSAIL MIT

Date: Monday, November 13, 2023

Time: 1:30 PM to 2:30 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: Patil/Kiva (Room 32-G449)

Event Type: Thesis Defense

Room Description:

Host: Daniela Rus, CSAIL MIT

Contact: Mieke Moran, 617-253-5817, mieke@csail.mit.edu

Relevant URL: https://youtu.be/5vD5nPL2w3Y

Speaker URL:

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: THESIS DEFENSE: Self-Assembling Modular Systems: Enhancing Efficiency and Accuracy in Robotic Lattice Assembly

ZOOM LINK: https://mit.zoom.us/j/96931618232

Abstract:
This thesis introduces a robotic system designed to progress the concept of modular 'industrial legos,' aiming to harness the advantages of modularity in new environments, like CNC machines and factory automation. This vision intersects with the domain of modular self-reconfigurable robotics and robotically-assembled structures. The work presented here extends from existing research and is particularly focused on three key areas: (1) the development of improved modular connectors between system elements, (2) the advancement of modular linear actuator designs that not only facilitate component rearrangement but also contribute to task performance, and (3) improved performance and efficiency of lattice-assembler robots. All topics are examined in depth, and their relevance to prior work is carefully outlined.
Central to any modular system is the design of connectors, which enable the modules to assemble into various configurations. Although extensive research has been conducted in this area, existing connector solutions often fall short in terms of repeatability, affordability, and mechanical strength. This thesis addresses these shortcomings by introducing a new connector that utilizes a kinematic coupling for high-repeatability and incorporates a simple, screw-based latching mechanism for strong mechanical connections. Detailed design specifications for these connectors are provided; these designs are both relatively simple to implement and robust enough to withstand the harsh conditions encountered in industrial settings.
In CNC and industrial automation, linear actuators often pair with structural elements to form gantry systems, allowing for precise 3D-positioning of tools and end effectors. While there are self-contained linear actuators that offer some level of modularity—often termed 'superficial modularity'—traditional designs generally lack the capacity for what might be called 'deep modularity.' This refers to the unique ability to seamlessly join two actuators of length L to create a single actuator of length 2L. To address this, the thesis presents a high-strength, extensible modular linear actuator built on magnetic lead screws, capable of general linear motion and of cooperating with other modular elements to rearrange modules.
Lastly, this thesis introduces an assembly-robot named Belty, equipped with proprioceptive actuators that feature large-diameter, high torque density brushless motors and minimal gearing. These actuators offer several advantages such as efficient motion, back-drivability, force sensing, and impact resilience. These features are valuable for assembly tasks within constrained spaces as they enable simplified assembly algorithms through contact-rich motion primitives (e.g, dragging parts, bumping into objects). Following the hardware discussion, assembly algorithms are detailed to demonstrate the system's ability to construct a diverse array of structures quickly and energy-efficiently.

Thesis Committee:
- Prof. Daniela Rus, Department of Electrical Engineering and Computer Science (Thesis Supervisor)
- Prof. Amos Winter, Department of Mechanical Engineering, (Committee Chair)
- Prof. Sangbae Kim, Department of Mechanical Engineering
- Dr. Beth Marcus, Amazon Robotics

Research Areas:

Impact Areas:

This event is not part of a series.

Created by Mieke Moran Email at Tuesday, November 07, 2023 at 4:11 PM.