Hi everyone!
I am using caffe to train a mobile robot to reach its target. Now that I finished the simulation environment, I have to build the real implementation.
The options for the hardware are two with the same budget:
- Desktop computer connects to mobile robot via raspberry or a generic laptop:
- pros: more powerful machine, faster tests in simulation, best GPU available is GTX 1080Ti
- cons: have to build a stable server-client connection, I was thinking about using DeepDetect or build my own communication with Qpid Messaging
- Laptop connects directly to robot:
- pros: less work needed by not developing a communication program across the network
- cons: less powerful (best GPU is GTX 1070), means also longer tests in both simulation and real environment
Personally, I would choose the first, but I would like to ask your opinion. Main questions I am asking myself are:
- Is the computational power difference big or not so much? Would I really feel it?
- Could be too much time consuming developing a Server-Client program?
Thank you for your attention.