One thing that sucked last winter was my kids would come down to my office ahead of needing a ride to various activities and I had forgotten to condition the car...sucked esp. when it's 5F outside and we had to leave immediately and so the car barely getting 1 min of preconditioning. Then I started training the kids to give me a 8-10 mins heads up to remind me to wake up and precondition the car, and if it wasn't plugged in, had to the run to garage and do plug in or estimate range loss preconditioning on battery. Kids proved to be super unreliable...I know first world problems.
So I figured I'd put my cs masters degree skills to some use.
Before scenario -
Unless you leave every day at the same time, you can use scheduled departure, which also cannot be done via the mobile app (why Tesla?) So you must use the mobile app to wake up the car in enough time to warm it up sufficiently, guessing how much time it will take to warm up accounting for outside temp and not garage temp (which in my case can be 40F off sometimes). Then turn climate control on and set the temperature maybe to conserve energy if not plugged in, set the seat heaters based on number of passengers and individual level preference, then remember to turn temps down/off after a few minutes especially if not plugged in. If you don't leave in time, remember to turn off conditioning otherwise get a message at 2 am (reminding you of your stupidity) that conditioning stopped because battery is at 20% (D'oh! and yes it's happened to me once). Conversely, on hot summer days, have preconditioning simply blast the AC in a closed car, without first venting windows to let out 150F air, which would be more efficient and save some battery or maybe conditioning to a nice 75F instead of 60F.
After scenario - Simply telling your voice assistant from your watch to "get the car ready" for x passengers in y amount of time (or at a future clock time). Let the AI decide based on multiple environment conditions, amount of drive time, the car state, and the previous learned actions what is the most efficient thing to do would be. Also, never ever needing to remember to turn off conditioning because it will do for you it based on power source and charge level.
Not sure how much energy is saved long run, but if we can have AI drive our car why can't AI get a car ideally ready using based on learning. Conditions hasn't changed much from ICE to EV, either blast heat at max or blast AC at max, So here is what I built, but first I had to build my own TFi equivalent to gather training data and then my own mobile assistant that could eventually make smart.
I will open source all this once i figure out how to package deployment on google cloud, as I don;t want to deal with getting passwords/tokens or dealing with takedowns.
So I figured I'd put my cs masters degree skills to some use.
Before scenario -
Unless you leave every day at the same time, you can use scheduled departure, which also cannot be done via the mobile app (why Tesla?) So you must use the mobile app to wake up the car in enough time to warm it up sufficiently, guessing how much time it will take to warm up accounting for outside temp and not garage temp (which in my case can be 40F off sometimes). Then turn climate control on and set the temperature maybe to conserve energy if not plugged in, set the seat heaters based on number of passengers and individual level preference, then remember to turn temps down/off after a few minutes especially if not plugged in. If you don't leave in time, remember to turn off conditioning otherwise get a message at 2 am (reminding you of your stupidity) that conditioning stopped because battery is at 20% (D'oh! and yes it's happened to me once). Conversely, on hot summer days, have preconditioning simply blast the AC in a closed car, without first venting windows to let out 150F air, which would be more efficient and save some battery or maybe conditioning to a nice 75F instead of 60F.
After scenario - Simply telling your voice assistant from your watch to "get the car ready" for x passengers in y amount of time (or at a future clock time). Let the AI decide based on multiple environment conditions, amount of drive time, the car state, and the previous learned actions what is the most efficient thing to do would be. Also, never ever needing to remember to turn off conditioning because it will do for you it based on power source and charge level.
Not sure how much energy is saved long run, but if we can have AI drive our car why can't AI get a car ideally ready using based on learning. Conditions hasn't changed much from ICE to EV, either blast heat at max or blast AC at max, So here is what I built, but first I had to build my own TFi equivalent to gather training data and then my own mobile assistant that could eventually make smart.
I will open source all this once i figure out how to package deployment on google cloud, as I don;t want to deal with getting passwords/tokens or dealing with takedowns.
