next up previous
Next: Sensing model and control Up: Problem Statement Previous: Problem Statement

Vehicle model

Figure 2: a) An agent's vehicle. b) The agent's windshield, or field-of-view
\begin{figure}\begin{center}
\begin{tabular}{ccc}
\epsfig{figure=figures/space...
...idth=0.4\textwidth} \\
(a) & & (b) \\
\end{tabular}\end{center}
\end{figure}

Consider a set of $ n$ agents, in which agent $ i$ is a point vehicle located at $ p_i = (p_{i1}, p_{i2})$ in the plane with orientation $ \psi_i$ (see Fig. 2(a)). Each vehicle moves as a Dubins car:

$\displaystyle \dot p_{i1} = v_i\cos\psi_i, \quad \dot p_{i2} = v_i\sin\psi_i, \quad \dot \psi_i = u_i,$ (1)

in which $ v_i$ is the forward speed, and the control is $ u_i \in \{-\omega_i, 0, \omega_i\}$ for some fixed $ \omega_i > 0$ . For fixed $ v_i$ , such a vehicle either moves along a straight line ($ u_i = 0$ ) or turns clockwise/counterclockwise ( $ u_i = \pm \omega_i$ ) along a circle with fixed radius. We use this fact later without further elaboration. Let $ X = SE(2)^n$ denote the state space, in which $ x \in X$ yields the position and orientation of all agents. Some of our results require that the agents are identical, which means $ v_i = v_j$ and $ \omega_i = \omega_j$ for all pairs, $ i,j$ , of agents.



Jingjin Yu 2011-01-18