Collaborative Control of Autonomous Swarms Under Communication and Resource Constraints
Doctoral Dissertation, Date: November 2006, Advisor: John S. Baras
Collaborative/cooperative control of a large group of autonomous vehicles has been received great attentions in recent years. With the rapid advances in sensing, communication, computation, and actuation capabilities, it is extremely appealing to control a large group of unmanned autonomous vehicles (UAVs) to perform dangerous or explorative tasks in various hazardous, unknown or remote environments. Possibilities of a broad range of applications by utilizing UAV swarms have been explored, for example, automated highway systems, mobile sensor networks in ocean resources exploration, spacecraft interferometry, satellite formations and robotic border patrol.
In such applications, traditional centralized control schemes are always prohibited primarily due to the high communication cost and the high computation cost in a large network of vehicles. In turn, the decentralized/distributed control schemes are preferred to achieve the trade off between the performance and the communication/compuation cost. In past decades, numerous decentralized/distributed control algorithms have been proposed in the literature. Among them, one approach, called bio-inspired approach, is extremely interesting and promising, which ”borrows” algorithms from nature by observing and understanding social animal’s swarming behaviors.
In this dissertation, we study a decentralized artificial potential function (APF) based approach which mimics bacteria foraging process. The deterministic potential based approach, however, suffers from the local minima entrapment dilemma, which motivate us to fix the ”flaw” that is naturally embedded. We propose an innovative decentralized stochastic approach based on the Markov Random Filed (MRF) theory, which traditionally used in statistical mechanics and in image processing. By modeling the local interactions as Gibbs potentials, the movements of vehicles are then decided using Gibbs sampler based simulated annealing (SA) algorithm.
A two-step sampling scheme is proposed to coordinate vehicle networks: in the first sampling step a vehicle is picked through a properly designed, configuration-dependent proposal distribution, and in the second sampling step the vehicle makes a move using the local characteristics of the Gibbs distribution. Convergence to the configuration(s) of global minimal potential is established theoretically and confirmed with simulations. In order to reduce the communication cost and the delay in the two-step sampling, a fully parallel sampling algorithm is studied and analyzed accordingly.
In practice the stochastic nature of the proposed algorithm might lead to high traveling cost and long maneuver time. To mitigate this problem, a hybrid algorithm is developed by combining the Gibbs sampler-based method with the deterministic gradient-flow method to gain the advantages of both approaches.
We also study the robustness of the Gibbs sampler based algorithm. The convergence properties are investigated under different types sensor errors including range-error and random-error. Some error bounds are derived to guarantee the convergence of the stochastic algorithm.
In order to integrate the Gibbs sampler based path planning algorithm in applications, a two-level scheme is proposed by combining high-level path planing and low-level vehicle motion control. The high-level path planing module mainly addresses the path generation. The low-level motion control module aims to follow the desired path by considering vehicle dynamics. A model predictive based (MPC) based motion control for car-like nonholonomic UAVs is investigated. Multiple control objectives, e.g., minimizing tracking error, avoiding actuator/state saturation, and minimizing control effort, are easily encoded in the objective function. Two numerical optimization approaches, gradient descendent approach and dynamic programming approach, are studied to strike the balance between computation time and complexity.