Robi改造计划更新---框架可以动了-CSDN博客

本文链接：https://blog.csdn.net/zister/article/details/53420681

落后原计划3个月，终于整个动作、听说、视觉可以整合在一起动了，在ROS框架下，并且整个由行为树（BT）控制，因为BT有很多随机的东西，所以整个机器人就显得很有意思了。

代码前几天已经更新在github，现在骨架可以动了，可以开始填肉了：

https://github.com/NickQian/Robi-Transform-Project/

TODO分4大块，列表如下:
1.a) servo action server now only support 1 servo command during server period. Need to support all servo at the same time.
1.b) Robi cannot move right now. need to add a balance car under the legs and let them move: controlled by BT.
1.c) integrate moveit! ?
2.a) sensors now are blank (except microphone and camera). Need to add sensors driver and integrate them into ROS to let them publish data.
2.b) integrate cartographer for SLAM
3.a) No object detection right now. Maybe "tensorbox" is a solution.
3.b) Object recognization using "inception" has not been integrated into ROS. also face recognization.
4.a) When speaking the microphone also capture the sound and record it. This is stupid. Need a way to improve this.
4.b) STT and TTS are not good. STT using Xunfei cloud and too slow. TTS result is not smooth. Needs better solution.
4.c) integrate "syntaxnet" to do some analyze?
4.d) dialog match now using AIML. need to imporove using RNN/LSTM things???