Research
Contingency-Aware Task Assignment and Scheduling for Multi-Agent Teams
We address the problem of task assignment and scheduling for human-robot teams to enable the efficient completion of complex tasks such as satellite assembly. To handle the uncertainty that arises from changing task requirements, potential failures, and delays, we must enable the human-robot team to work together effectively. Our work makes two main contributions: (1) we use a multi-agent concurrent MDP framework to account for the complex interaction of uncertainty that stems from the tasks and the agents, and (2) we use Mixed Integer Linear Programs and contingency sampling to approximate action values for task assignment. Our algorithm is computationally efficient while making optimal task assignments compared to a value iteration baseline.
Human-Guided Goal Assignment to Effectively Manage Workload for a Smart Robotic Assistant
In human-robot teams, managing robot workloads is critical for efficient team operation. Overloading robots with work can lead to missed deadlines and additional work for humans. This paper presents a framework for a robot to assess its own workload based on an initial goal assignment. The robot generates task and motion plans and computes the probability of missing deadlines due to potential delays in task execution. A branch and bound-based search algorithm is used to generate task and motion plans by minimizing task execution effort. The robot then presents a diverse set of task and motion plans to humans to offer multiple options. Humans can approve a plan or provide guidance to reduce the workload by relaxing deadlines or removing assigned goals. This framework enables the robot to assess its workload and present solutions to humans in a collaborative and efficient manner.
Mobile Manipulator System for Accurate and Efficient Spraying on Large Surfaces
Our team has developed a mobile manipulator system for accurate and efficient spraying on large surfaces, which is essential in many industrial applications. Our system generates a plan automatically based on the given spraying task, enabling the robot to spray on the entire surface. We have used self-supervised batch learning to reduce the number of experiments needed to create a model of the spray tool. In addition, we have created a mobile base placement planner that determines the minimum base locations required to carry out the spraying task. Our image-based perception pipeline allows the robot to characterize spraying error. We have experimentally verified these algorithms by having a mobile manipulator spray paint a large mural.
SMAC: Symbiotic Multi-Agent Construction
This developed an innovative concept for a distributed platform that enables autonomous 3D construction using two types of agents working together in a coordinated manner. The platform includes a set of smart construction blocks that are capable of planning and monitoring their own status as well as the progress of construction. Additionally, we have a team of inchworm-inspired builder robots designed to navigate and modify the 3D structure, following the guidance of the smart blocks. We have designed the hardware and created algorithms for navigation and construction that support a wide range of 3D structures.