Fix a issue when test multi-round in simulation

README.md

@@ -10,9 +10,9 @@ Some realworld experiment: [YouTube](https://youtu.be/LHvtbKmTwvE), [bilibili](h
 ## Introduction:
 We propose **a learning-based planner for autonomous navigation in obstacle-dense environments** which intergrats (i) perception and mapping, (ii) front-end path searching, and (iii) back-end optimization of classical methods into a single network. 
 
-Considering the multi-modal nature of the navigation problem and to avoid local minima around initial values, our approach adopts a set of motion primitives as anchor to cover the searching space, and predicts the offsets and scores of primitives for further improvement (like the one-stage object detector YOLO). 
+**Learning-based Planner:** Considering the multi-modal nature of the navigation problem and to avoid local minima around initial values, our approach adopts a set of motion primitives as anchor to cover the searching space, and predicts the offsets and scores of primitives for further improvement (like the one-stage object detector YOLO). 
 
-Compared to giving expert demonstrations for imitation in imitation learning or exploring by trial-and-error in reinforcement learning, we directly back-propagate the numerical gradient (e.g. from ESDF) to the weights of neural network in the training process, which is realistic, accurate, and timely.
+**Training Strategy:** Compared to giving expert demonstrations for imitation in imitation learning or exploring by trial-and-error in reinforcement learning, we directly back-propagate the numerical gradient (e.g. from ESDF) to the weights of neural network in the training process, which is realistic, accurate, and timely.
 
 <table>
     <tr>