General

Tesla Forums are now read only. To continue the conversation with the Tesla community visit engage.tesla.com.

Terrain Aware Speed Control

What if users were able to input a speed fluctuation value that autopilot used to increase / decrease speed on hill ascent / descent? For example, I input a speed fluctuation value of 5 MPH and autopilot allows the vehicle to lose 5 MPH on hill ascent and gain 10 MPH on hill descent. I live in StL MO and there are lots of small hills everywhere I go. I've used this approach manually around town and can greatly increase efficiency of my M3. I imagine software could do an even better job with even less speed fluctuation. Thoughts?

Comments

  • Since EVs have regen, it does not really matter as long as all the braking down hill is done with regen. Maintaining speed will get you the most.
  • Doing what the OP suggests at say 35mph speed limit would do better. Regen loses about 15% in conversion losses and the speed increase from 35 to 45mph would not even come close to negating that due to the additional aerodynamic drag losses. I have had good success with even just staying at the speed limit and letting it gain 5 mph on downhills and then let it bleed back off here in Northern Virginia.

    Unfortunately, I think the first step would be to make AP a little more smooth as currently it is not the most efficient driving mechanism. Even normal driving for me it better than AP from an efficiency standpoint if for nothing else slowing down for an upcoming red light. AP waits too long to start slowing down. Also it should rubber band a bit more with traffic...within reason of course.

    I know the little hills well in St. Louis, grew up in Des Peres and I visit a couple times a year.
  • > @derotam said:
    > I have had good success with even just staying at the speed limit and letting it gain 5 mph on downhills and then let it bleed back off here in Northern Virginia.

    Yep exactly. If the calculation of speed gain on descent and speed bleed on ascent was done by the CPU taking current grade as well as upcoming grade into consideration I think we could see some serious efficiency gains. AP can get pretty aggressive with acceleration on ascent. I assume currently AP does not consider grade at all, and will accelerate when it notices speed reducing on ascent which is the most inefficient way to operate.

    I am just north west of Des Peres. Easy living.
  • A quick qualifier - the range of my standard range rwd M3 is more than enough and I love to stretch her legs quite often. However once and a while I also enjoy eking out every last bit of mileage / kwh. Both approaches are quite entertaining. All depends on my mood :).
  • I agree.

    Time to go eat dinner and eat some Ted Drewes I brought back with me when I was there a couple weeks ago! With the dry ice and cooler it all stayed frozen for the 16 hour drive!
  • > @derotam said:
    > ...
    > Unfortunately, I think the first step would be to make AP a little more smooth as currently it is not the most efficient driving mechanism. Even normal driving for me it better than AP from an efficiency standpoint if for nothing else slowing down for an upcoming red light. AP waits too long to start slowing down. Also it should rubber band a bit more with traffic...within reason of course.
    >

    Exactly. With all of the AI running on this car, it really should anticipate lights and other road obstructions better. Only the last few feet should involve friction brakes.
  • It can, its all about programming in the desired behavior.

    With brake hold you don't even need friction brakes for the last few feet.
  • Not on my 2018S. Once it gets down to a very slow speed it rolls. The brakes are needed. Brake hold just used the ABS valving to hold pressure against the slave cylinders.
  • Sorry, my comment was in reference to the hold mode on the 3. On the 3, hold mode uses regen or negative power to the motor all the way to 0mph and then at 0mph engages physical brakes to hold.
  • > @beammeupscotty said:
    > What if users were able to input a speed fluctuation value that autopilot used to increase / decrease speed on hill ascent / descent? For example, I input a speed fluctuation value of 5 MPH and autopilot allows the vehicle to lose 5 MPH on hill ascent and gain 10 MPH on hill descent. I live in StL MO and there are lots of small hills everywhere I go. I've used this approach manually around town and can greatly increase efficiency of my M3. I imagine software could do an even better job with even less speed fluctuation. Thoughts?

    Reducing speed uphill looks like a starter for road rage behind and increasing speed downhill looks like a potential speeding ticket.
  • > @derotam said:
    > Doing what the OP suggests at say 35mph speed limit would do better. Regen loses about 15% in conversion losses and the speed increase from 35 to 45mph would not even come close to negating that due to the additional aerodynamic drag losses. I have had good success with even just staying at the speed limit and letting it gain 5 mph on downhills and then let it bleed back off here in Northern Virginia.
    >
    > Unfortunately, I think the first step would be to make AP a little more smooth as currently it is not the most efficient driving mechanism. Even normal driving for me it better than AP from an efficiency standpoint if for nothing else slowing down for an upcoming red light. AP waits too long to start slowing down. Also it should rubber band a bit more with traffic...within reason of course.
    >
    > I know the little hills well in St. Louis, grew up in Des Peres and I visit a couple times a year.

    I agree with all of this. Especially the second paragraph.
  • I think we all agree that AP is simply not set up for efficiency at this point. Perhaps the roadmap is to get FSD online then address efficiency. I would love to see the addition of grade aware speed fluctuation. Or perhaps a drive mode for AP similar to chill / sport / etc. If a user selects chill AP errors on the side of efficiency when deciding when and how much to accelerate / break.
  • I doubt that we'll see Tesla setting the AP algorithms to optimize for efficiency in any way being described above. They very much don't want Tesla's to get the bad reputation that Prius's got by drivers annoying other drivers on the road with their hypermiling antics that disrupt traffic flow.
    Tesla's AP is set to drive with the flow of traffic. If you want to reduce your carbon footprint, they will sell you more solar panels.
  • @"grins.va" as with everything, it is situation dependent and should only be used in "appropriate" scenarios.

    Lots of people naturally slow down a bit going uphill just because they are unable to appropriately adjust for the hill.
  • @"Earl and Nagin 08 RDS 359" @"grins.va" It should be up to the user in my opinion. Or make AP traffic and grade aware. The data for terrain is out there. Almost every bike route planner tells you incline / decline. Why not incorporate this into AP? I'd love to have a hypermile AP mode. If there is traffic around it sticks with traffic flow. No traffic, Sunday afternoon, majestic American backroad - slow and low baby! And maybe with the efficiency gain from smart, efficient AP I can even roll the windows down.
  • AP is already traffic and grade aware both uphill and down. It maintains the speed the operator has set based on following distance. You can hypermile using your right foot.
  • In my experience with 2020 M3 AP is really only aware of the traffic 200-300 feet ahead. As a result it drives like a teenager speeding up to the limit set until it has to slow down quickly because of slower traffic ahead. This and other experiences suggest to me at least that the self driving software can not plan ahead because it can not see ahead.
Sign In or Register to comment.