1.How to convert the waypoint output of the model into the executable linear speed and angular speed of the car chassis?
my conversion process is as follows where DT represents the reciprocal of the frequency of sending messages using ROS, and dx, dy represents the model output wayponts.
v = dx / DT; w = np.arctan(dy/dx) / DT
2、I tested the pre-training model, and it seems that the obstacle avoidance effect is not very good. When the car approaches the obstacle, it still output a large linear speed.
I'm not sure where the problem is, looking forward to your reply, thanks!
Thank you for your excellent work, and I am very interested in your work，When I tried to deploy the model in your paper Learning Composable Behavior Embeddings for Long-horizon Visual Navigation to my real car, I encountered the following problems：
1.How to convert the waypoint output of the model into the executable linear speed and angular speed of the car chassis? my conversion process is as follows where DT represents the reciprocal of the frequency of sending messages using ROS, and dx, dy represents the model output wayponts.
v = dx / DT; w = np.arctan(dy/dx) / DT2、I tested the pre-training model, and it seems that the obstacle avoidance effect is not very good. When the car approaches the obstacle, it still output a large linear speed.I'm not sure where the problem is, looking forward to your reply, thanks!