Tracking evaluation

The tracking performance is evaluated through a computer simulation with various initial states of a mobile robot. It is assumed that a target human is walking parallel to the positive direction of the X-axis. The initial state vector of the human $(-1500,1000,0)$ and $\theta_{sh}$ is set to $180^{\circ}$. The trajectories of the human and the mobile robot are shown in Fig.7(a). Fig.7(b) shows how long it takes before convergence on the trajectory of the human is achieved. $\alpha$, $\beta$, and $\gamma$ in Fig.7(a) correspond with Fig.7(b). Although the mobile robot starts from various initial states, it is able to track the target and converge on the human trajectory in a reasonable length of time. From the results, it can be concluded that the proposed control law is valid.

(a)

(b)

図 7: Convergence evaluation

According to changes in the parameters, the mobile robot's trajectory changes, as shown in Fig.8. In this simulation, the target human makes a right-angled turn, as shown in figure. $k_2$ and $k_4$, which are related to the spring bending, are set constant through all of this simulation. Fig.8(a) is the trajectory of the mobile robot when $k_1$ is changed and $k_3$ is constant. A small $k_1$ makes the performance on the right-angled turn worse. On the other hand, the mobile robot strays from the human trajectory with a large $k_1$. If $k_1$ increases, the trajectory of the robot after turning becomes oscillatory. In Fig.8(b), both $k_1$ and $k_3$ are changed. The performance as a result of a right-angled turn is improved by adequately setting up $k_1$ and $k_3$. The adequate parameters are tuned experientially by performing this simulation so that the mobile robot does not stray from the trajectory of the human and the motion of the mobile robot does not oscillate.

(a)

(b)

図 8: Changes in the parameters