A Scalable Control Framework for Boundary Coverage and Cooperative Manipulation by Robotic Swarms

Project: Research project

Description

Overview: Advances in computing, sensing, actuation, and control technologies are enabling the development of swarm robotic systems comprised of hundreds or thousands of robots that execute tasks as a collective with little to no supervision. These systems would offer efficient automated solutions for tasks that require high degrees of parallelism, redundancy, and adaptability. Key requirements in controlling such large robot populations include the accurate prediction of group behaviors and the synthesis of control policies that are scalable with the population size, accommodate the robotic platform limitations, and provably produce a desired macroscopic outcome. The proposed work will develop a rigorous approach to robotic swarm control that satisfies these requirements and applies to scenarios in which the robots must (a) allocate themselves around the perimeters of regions or objects in their environment, and (b) cooperatively manipulate and transport payloads to which they have allocated. The control approach does not require global position information, inter-robot communication, or prior information about the regions or objects. The approach will employ stochastic, deterministic, and hybrid stochastic-deterministic models at different levels of abstraction that describe the robots roles, task transitions, motion, and manipulation dynamics. Control and optimization techniques will be applied to macroscopic models of the swarm dynamics in order to synthesize robot control policies that produce target allocation and manipulation behaviors using only local sensing and common broadcast information. These techniques will include feedback control of linear and multi-affine dynamical systems, including piecewise-multi-affine hybrid systems on polytopes; convex optimization; and stochastic optimization methods. Control strategies for cooperative manipulation will be developed in part from models of desert ant group retrieval that are based on previously collected experimental data using ant force sensors. The control approaches will be validated with computer simulations at varying levels of fidelity and through experiments on a testbed with small differential-drive robots. Robot grippers with three degrees of freedom will be designed and implemented for use in the cooperative manipulation experiments.
StatusFinished
Effective start/end date8/1/147/31/17

Funding

  • National Science Foundation (NSF): $260,000.00

Fingerprint

Robotics
Robots
Grippers
Convex optimization
Stochastic models
Testbeds
Hybrid systems
Feedback control
Redundancy
Dynamical systems
Experiments
Communication
Sensors
Computer simulation