Collaborative Modular Robot Teammates for Shipboard Inspection and Maintenance
Abstract
Inspection, maintenance and emergency response are three of the most important shipboard tasks since vessels provide human living on a fuel storage farm and a power plant with high voltage. Vessels are structured complicatedly and equipped with numerous devices and tools. Many inspection and maintenance tasks are currently conducted by humans, so it is imperative that these tasks be automated for the maximization of safety and the minimization of workload. The fact that these tasks consist of complex human operations has given rise to need for robots which have human-like intelligent perception and flexible mobility/manipulability.The goal of this project is to develop autonomous robot teammates that perform shipboard inspection and maintenance tasks collaboratively. To achieve the goal, the technical objectives defined for the project are: Robots: Design and develop low-cost robots that consist of modular components and can form a variety of heterogeneous robot team; and Collaborative intelligence: Develop a scheme of the heterogeneous robot team that collaboratively (with intelligence) move, manipulate and complete shipboard tasks including inspection, maintenance and emergency response in shipboard environments. The robots to develop consist of modules including head, upper body, lower body, arms, legs, hands and feet/wheels. Different modules will be plugged and played to form a robot with different capabilities. The modular robot will be centrally controlled to minimize the number of electric components. Custom data/power connectors and hub will be then developed to plug and playmodules through the connection of one cable.The proposed collaborative intelligence approach is characterized by the four unique in-house technologies. The primary characteristic of the proposed approach is autonomous position/force controlled manipulation using multi-stage target observation and whole body control. To handle uncertainties of complex environments, perception implements the multi-stage probabilistic target observation technique, which maximizes the detection events while making the manipulation problem Gaussian. High-precision manipulation aimed in the project extends the multi-contact whole body control, which was proposed by the Principal Investigator (PI). The second characteristic is Bayesian Search, Tracking, Localization and Mapping (STLAM), which is anestimation and control framework of autonomy proposed by the PI. The strength of the Bayesian STLAM comes from the consideration of uncertainties of both the robot and target. The third characteristic is human supervision with multi-level autonomy, which is achieved by adding robot sensing capability. The last characteristic is multi-robot collaboration using belief fusion, which improves mission completion through an effective multi-robot collaboration scheme. Theproposed approach including the multi-contact wer scientists at University of Maryland.The developed technologies will be ultimately evaluated and demonstrated in shipboard-like environments. Tasks to demonstrate are all for shipboard inspection, maintenance and emergency response such as inspection/maintenance tool picking, large valve rotation, heavy object carrying,cart pushing or hose pulling. Demonstration is an important element of this project as shipboard inspection, maintenance and emergency response are highly practical tasks. Many naval inspection and maintenance tasks are currently conducted by humans, so replacing humans by robots with improved autonomy will contribute to the increase of safety and the decrease ofworkload and make an impact on the DoD capabilities.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 08, 2020
- Source ID
- N000142012468
Entities
People
- Tomonari Furukawa
Organizations
- Office of Naval Research
- United States Navy
- University of Virginia