Limitations of Employing Undirected Information Flow Graphs for the Maintenance of Rigid Vehicular Formations

Author(s):  
Swaroop Darbha ◽  
Prabhakar R. Pagilla

In this paper, we consider the problem of decentralized control of a collection of homogeneous vehicles trying to maintain a rigid formation. Each vehicle attempts to maintain a specified constant safe distance from its adjacent vehicles in the collection. We consider an identical structure for each decentralized controller so that it is simpler from an implementation viewpoint as it does not depend on collection size or vehicle indices. We call a vehicle B adjacent to vehicle A if the relative position of vehicle B is known to vehicle A either by communication or by sensing. In this paper, we only consider undirected information flow graphs, i.e., graphs where vehicle A is adjacent to vehicle B if and only if vehicle B is adjacent to vehicle A. We consider a point mass model for each vehicle and assume the actuation transfer function, which relates the control input to the force supplied to the vehicle, to be a strictly proper rational transfer function. It is known that spacing errors due to disturbances amplify if the reference vehicle information is not available to Ω(n) vehicles, n being the size of the collection [1]. In this paper, we generalize this result to show the following: If there are two or more vehicles in the collection that are adjacent to Ω(n) vehicles, then there is a critical size N* so that the motion of the collection will be unstable if the size of the collection exceeds N*. Practical issues of fault tolerance indicate that there be at least two vehicles that are adjacent to Ω(n) vehicles in the collection. We also further show that the use of a kinematic vehicle model for analysis of disturbance propagation yields results which may not agree with what is observed in practice and hence are inappropriate.

2005 ◽  
Vol 24 (2) ◽  
pp. 125-134
Author(s):  
Manabu Kosaka ◽  
Hiroshi Uda ◽  
Eiichi Bamba ◽  
Hiroshi Shibata

In this paper, we propose a deterministic off-line identification method performed by using input and output data with a constant steady state output response such as a step response that causes noise or vibration from a mechanical system at the moment when it is applied but they are attenuated asymptotically. The method can directly acquire any order of reduced model without knowing the real order of a plant, in such a way that the intermediate parameters are uniquely determined so as to be orthogonal with respect to 0 ∼ N-tuple integral values of output error and irrelevant to the unmodelled dynamics. From the intermediate parameters, the coefficients of a rational transfer function are calculated. In consequence, the method can be executed for any plant without knowing or estimating its order at the beginning. The effectiveness of the method is illustrated by numerical simulations and also by applying it to a 2-mass system.


2012 ◽  
Vol 18 (5) ◽  
pp. 425-436 ◽  
Author(s):  
Rubens H. Korogui ◽  
André R. Fioravanti ◽  
José C. Geromel

Sign in / Sign up

Export Citation Format

Share Document