There's actually two ways to do that.

The most common is indirect. The robotic prothesis may have a termometer of some sort, which can either show the actual temperature in a visor somewhere or "raise a flag" (blink a led, sound a beep, vibrate a little, etc) indicating it's over certain temperature (usually what would be considered too hot). It can also simply transmit the same temperature to another area that's still functional, but this tech needs improving as any delay could cause harm if someone get's something to hot to drink for example.

The other way requires to actually have that termal signal converted into electrical signal of pre-determined current that goes into an electrode to stimulate the corresponding area of the brain.

This is easier said than done, there is some devices in development that can feel pressure like human skin. Google robotic skin and you'll find out that the solutions are fairly low tech and pretty functional when it come to the robotics, but when trying to apply that to a human protesis it becames the problem as of right now, there is no correct way to stimulate the brain.

Any neurosurgeon can stimulate the brain with electrical current. The problem is when we want to leave it up to a machine to do that. Firstly, because when neurosurgeons do that, usually the pacient has it's head opened while he can carefully manuever metal needle-like electrode that needs to touch the brain in the correct spot (which can be pretty deep), so a machine would need to do that with the skull closed and don't mess with those electrodes or risk damaging the brain. Secondly, although anatomically close, we're all different up there, so it's not practical to simple plan the machine to cause stimuli in one area, as that area may not respond to the same thing in you and me. This is specially problematic when we consider cases like the one in this news, victims of stroke may naturally come to use healthy areas of the brain to work in place of the areas that were affected.

And even if we get the plans right, in the end, the protesis would only work for one person and everything welearn from building it may as well be worth nothing to make the next one. So, as you can guess, there a lot of money restriction to this kind of research too as, right now, there's absolutely no foreview for mass production or application.

So, the tech is pretty much here. But making protesis with feedback practical may take a couple decades more. More integrated robotic protesis will probably came in much sooner.