My name is Ravin Menghani. I had couple of questions regarding Go1 and Go2. Currently my university Hochschule Anhalt, is having Go1 robot. Can you please tell me what is the difference between Go1 and Go2. As we recently saw some videos of Go2, but we already have Go1, so can you please help us to know more about Go2 and its comparison with Go1.
The Unitree Go2’s software is much better than the Unitree Go1’s software especially the high level control of the robot, but unlike the Go1, you don’t have access to the main PC. Also, it has a single PC 40 TOPS/100 TOPS for processing that comes with the edu version. If you plan to use the robot with reinforcement learning for low-level control i.e. controlling each joint separately, the Go1 should do the job, if you want to develop some application without worrying about low-level control, I would suggest the Go2.
In terms of specifications, I would suggest going through our qre_docs in GO2 manual subsection.
I had couple of questions, in the manual it is mentioned we have mid360 Lidar on Go2, however the demonstration video given there is of 2D slam, and it is written that we can enable mid360, but where is it mounted if the front lidar is a 2D lidar.
Also regarding accessing main PC, if we do not have access to main PC in edu version, how can we upload our own ros packages in the robot and run, or if we mount our own camera or lidar, how can we add that in the urdf of the robot. Also if we want to have our own modes, then how can we use the navigation stake and everything if we do not have access to the main PC the way we had in Go1 using SSH. Like is it possible in Go2 to do all these things? I am little confused here with low-level and high-level applications.
The mid 360 is a separate lidar that is mounted on the back of the GO2. it does not come by default as the one in the front.
The 2D slam works using the go2 head’s lidar albeit, it is not reliable. If you have another lidar like the mid360, it would be preferable. Also, the 2D scan is compressing the 3D point cloud from a certain height into 2D.
The front is a 3D lidar more specifically Unitree 4D L1 lidar.
You have access to a nvidia dock that comes in the back. I think you have options of 40 TOPS and 100 TOPS. In these nvidia boards you receive all info similar to go1 in DDS topics and moreso if you are using the qre_go2 repository which is for our customers.
All accessible in the Nvidia PC
You do not need access to the main PC for navigation stack or for low-level individual joint control.
Ok, so we have robosense Hellious 16P lidar with us in Go1, and we already had all the setups in it like drives and topics, so for Go2 we have mid360 in the same way as per what I understood from manual but, can we get everything we had in our Go1 like drives and topics for Hellious-16P somehow?
So, basically 2D we get from the inbuilt 3D LiDAR and compressing it, so we can access the 3D data somehow I believe, but just a question out of curiosity, that why are we compressing 3D lidar to 2D, isn’t 3D LiDAR better or 2D?
Oh, so we have a nvidia board same as Go1, but Go1 is having 3 nvidia boards and Go2 we have only 1 board right? Also how can I access qre_go2 repository I can access it only after we purchase Go2 or can I access it somehow now? Because I also want access to Go1’s repository, it would be great of you could provide me link and access for Go1’s repository my github ID: Rocky170500
Also, we wanted to know, can we transfer everything we have in our Go1 to Go2? As we done couple of projects and we don’t want to loss them, so it would be great if we could somehow transfer everything. As we also have a robot arm installed on the Go1 robot, with ZED camera and Robosense LiDAR as attachments.
3D is should be better but computationally its more expensive and causes lag in the Nav2 server while in most cases not providing more robust localization. Also the 3D localization is not available for Nav2’s cost maps which we can use for forbidden zones and/or speed zones.
Yes, you gain access to the Go2 after purchase, for Go1 please email support@mybotshop.de with your purchase id.
Hardware can be transferred, for software, you would have to port it to ROS2.
However, if I convert my rich 3D SLAM map to 2D map and then use it for Nav2, then will it work? as in input for nav2 will be a global map which is at the end of the day 2D and runtime as per the real-time requirement can be also 2D localization. Will Go2 and Nav2 used in Go2 support this Idea?
I will ask for the purchase ID and mail on this email address
Ok, so currently Go1 uses ROS1 melodic, and Go2 uses ROS2, so drives and everything of Go1 can not be directly transferred to Go2 and there is difference in ROS versions and I believe there is also some limitations of some drives that they support only ROS1 and or something similar, so I am not able to understand how can we do porting, can you please explain a bit in detail ?
Almost everything in ROS1 is now available in ROS2 Humble, Jazzy and above, unfortunately Foxy has many things missing and there will be issues in porting.
Mostly its changing sytanx and library names, if you’ve written the ros1_drivers and ros1_control yourself for the robot or a sensor than that will definitely need to be re written.
I want to ask you that can we have a meeting, so that my Professor and University can also be included as we are interested in buying Go2 because we want this self charging function which is not possible in Go1. So it would be great if we all could have a meeting by next week so that we can finalize everything and also how can we change from Go1 to Go2.
Unfortunately, neither the Go2 or the docking station is waterproof
No the Go1 is not water proof any sense. The only quadruped fully water proof is the B1 or B2.
Stepping on it wont harm you, if you put two exposed limbs between the plates then some voltage may pass by definitely. I will just incase confirm with Unitree.