Today in a tweet, Elon Musk eluded to hardware and software redundancy coming.
He then further clarified a question about how this will work:
A few questions come up as I think about this:
1. Will each SoC run identical versions of the OS and NN?
- If so, will this potentially open them up to simultaneous crashes due to the identical bugs?
- If not, could this provide additional resiliency?
- Will Tesla possibly operate two NN models simultaneously?
- How will Tesla handle conflicting data and/or split brain?
2. Since redundancy is going to play a more significant role as autonomy increases, will Tesla eventually install additional radar sensor, or is the plan to fall-back on cameras?
3. How will redundancy affect uptime targets? Will we be able to achieve five-9's of reliability?
He then further clarified a question about how this will work:
A few questions come up as I think about this:
1. Will each SoC run identical versions of the OS and NN?
- If so, will this potentially open them up to simultaneous crashes due to the identical bugs?
- If not, could this provide additional resiliency?
- Will Tesla possibly operate two NN models simultaneously?
- How will Tesla handle conflicting data and/or split brain?
2. Since redundancy is going to play a more significant role as autonomy increases, will Tesla eventually install additional radar sensor, or is the plan to fall-back on cameras?
3. How will redundancy affect uptime targets? Will we be able to achieve five-9's of reliability?
