Tesla Smart Summon needs to get a lot smarter

Just over a week ago, I purchased the Full Self Driving package for my Tesla Model 3. My car now has some amazing new capabilities that far surpass anything...

Just over a week ago, I purchased the Full Self Driving package for my Tesla Model 3. My car now has some amazing new capabilities that far surpass anything else on the road.

While that’s amazing, it is worth talking about the fairly substantial to-do-list on some of the FSD features. The one I want to focus on today is Smart Summon.

When you buy FSD, a new option, Summon, appears in your Tesla mobile app shows up. Standard Summon allows you to move the car back and forward which in a practical sense would allow you to move your car out of a tight parking space and enter the vehicle without damaging your doors.

The other option is Smart Summon. This feature shows a top-down view of your car and has two modes to route the car. The first is to navigate to a location (within a car park), which would theoretically be used to send the car to go park. The second option is to ‘Come to me’ which would be used to have the car navigate from its parking position to get you, say if it was raining for example.

The feature is still in beta, much like many of the FSD features, but without feedback and pressure from Tesla owners, I don’t feel the potentially great feature will get the development priority it deserves.

The reality is that Summon’s routing through a parking lot has major issues. While Tesla’s on-road navigation is fantastic, it’s more localised routing from point A to point B in a parking lot, is just straight up bad right now.

Naturally the moving Tesla has all the necessary radar, computer vision and sensor array to avoid pedestrians and objects like cars, trolleys etc. So my issue is definitely not with the safety of the system, if anything it’s too cautious.

The most obvious path planning consideration that’s missing is collision avoidance with the parking bays. The projected path between the start and endpoint should never overlap the parking bay, but it almost always does.

It’s really hard to understand how the current system is so bad, with the most obvious choice being a straight line (shortest distance), but often the app shows Smart Summon wants to take a strange, incredibly inefficient route, sometimes in the complete opposite direction to what I would drive.

I started testing Smart Summon in a completely empty car park until I understood the abilities and limitations of the feature. As you can see it had no problems driving over the white lines of the car parking bays.

Next test was in a quiet carpark, but did have other vehicles. This was particularly challenging given the high gutters and limited angles that would get the vehicle out of the space (rims were at risk if this went badly).

The car reversed, then gave up with an obstacle (parked car) behind and the gutter in front. To be honest, this was difficult, even for a human to navigate, with a car beside as well, limited the potential to reverse at an angle.

Next was in an underground car park, which I was impressed by the range at which the feature worked and that it worked at all in a concrete bunker.

After requesting the vehicle come to me, it moved out of the park and started heading towards me. It was working, but slowly, too slowly. An approaching car made me give up on the test and run over to the car, jump in and drive away.

Today was probably my 6th attempt in a week to use Smart Summon and I thought I’d set up the easiest test yet. Come and pick me up, in what was virtually a straight path with no obstacles in the way. The system failed badly.

It seems in this test, my Model 3 wanted to 4WD it through a garden.

Having used way points when flying a DJI drone, it made me wonder if we couldn’t get an app update that allowed us to plot the route and help train the AI to do a better job.

Once path planning becomes more reliable, the system then has to deal with multi-story carparks. Given the current system gives up on anything greater than a 10 degree gradient, there’s much work to be done here.

I think it’ll be a long time before we progress to the point where we feel comfortable using summon outside the line of sight, but it really could be one of the car’s greatest features.

Ordering a Tesla Model S, 3, X or Y
Please use our unique referral link for free Supercharging – https://ts.la/jason45054

The concept of Smart Summon is great, drive up to the door of a shopping center, get out and the car goes to park itself. When you’re done shopping, the car comes to pick you up. In the dry this is fancy, in the wet, this could eliminate your need for an umbrella.

Right now, with how many issues Smart Summon has, it’s a little more than a party trick to show friends. I fear that many Tesla owners try and give up and the system won’t learn as it does with Autopilot and the millions of km per month that are feeding into Tesla’s AI to improve it.

It really is something special to drive your Tesla like an RC car (up to 10km/hr) with nobody in it. That’s amazingly unique and cool and fun, but that fun would be made so much better if the system actually worked.

What is great is that those of us who purchased the Model 3 when HomeLink was included, now get a button in the Controls page of the Tesla app. This means you can open or close your garage door on demand and if you use Summon to reverse your car out, it’ll open the garage door before trying to drive out.. smart.

My advice is don’t buy FSD to get Smart Summon, it’s definitely not worth $8,500 for this feature. What FSD represents is much bigger than this single feature though, it’s about what’s coming and that’s hard to put a price on.

Categories
Tesla

Creator of techAU, Jason has spent the dozen+ years covering technology in Australia and around the world. Bringing a background in multimedia and passion for technology to the job, Cartwright delivers detailed product reviews, event coverage and industry news on a daily basis.
No Comment

Leave a Reply

Related Posts