How XPENG Is Redefining the Seat as Something That Senses, Decides, and Acts

A sketch of the 128-degree "zero-gravity mode." The footrest lifts as weight spreads across the back, waist, hips, and legs.

The backrest reclines to about 128 degrees, the footrest rises — XPENG's "zero-gravity mode" is the easiest way to picture where seats are headed. But the company's real bet isn't on angle, or on how many massage points a seat has. It's on whether a seat can read a passenger's build, posture, pressure distribution, and vital signs, then decide what to do about it on its own — moving from a seat that acts on command to one that notices a change in the body before anyone says a word. The part of the car that stays in contact with the body longest is starting to change.

By Sang Min Han _ han@autoelectronics.co.kr
한글로보기






The backrest tips back. The footrest rises, and body weight spreads across the back, waist, hips, and legs. Ma Jia, XPENG's head of interior and seat engineering, presented this near-128-degree recline — meant to mimic the feeling of floating in space — as one scene from the seat's future.
The angle is what catches the eye first. But the number itself wasn't the point Ma Jia wanted to make. Car seats have spent decades adding adjustment functions: fore-and-aft sliding and backrest tilt gave way to multi-directional power adjustment, memory settings, heating, ventilation, massage, powered lumbar support, and leg rests. Features once reserved for six-figure luxury cars now come standard on RMB 100,000-class (roughly $14,000) vehicles in China.
As these features spread quickly, the old way of standing out ran out of room. Ma Jia used massage as his example.
"You could keep adding massage points — sixteen, eighteen, twenty-two — but there's only so much physical space inside a seat."
Every extra adjustment axis, every added massage point, buys less differentiation than the last. What Ma Jia proposed instead was changing how the seat adapts to a person in the first place.
"Going forward, the seat won't be passive hardware waiting for commands. It will become an active, intelligent system that understands a user's state and intent on its own and responds before being asked."
Height, weight, and build differ from person to person. The same person doesn't sit the same way all day, either — posture shifts, and pressure builds up in different places, depending on how long someone's been driving, the road surface, and fatigue. Seats have always left it to the driver to fix this: move the backrest when it's uncomfortable, adjust the lumbar support when the back aches, hit the massage button when tired.
In the future Ma Jia describes, that order reverses. The seat notices the change in the body before a person feels the discomfort and reaches for a button.


 

Where It Touches the Body Longest

Cameras watch the passenger. Microphones pick up voices. Radar and ultrasonic sensors track people and movement inside the cabin. The seat has one thing none of those have: it stays in contact with the largest part of the body for as long as someone is in the car.
That contact point carries more information than it might seem. Pressure sensors can tell where weight is distributed, whether a passenger has slumped to one side, how posture shifts over time. Add in the car's cameras, radar, and biometric sensors, and the system can also read facial expressions, gaze direction, heart rate, and breathing.
The question isn't how much data this adds up to — it's what gets done with it. If a passenger leans to one side, the seat can shift support to match the new pressure distribution. If the same posture holds too long, it can adjust the backrest or cushion before fatigue sets in. Massage, instead of running a fixed program, can target wherever fatigue has built up, changing position, intensity, and rhythm accordingly.
None of this needs to stay inside the seat, either. If heart rate or breathing shifts away from the norm, the seat can alert the passenger, or adjust climate control, lighting, music, and massage together.
Heating, ventilation, massage, lumbar support — these functions already exist. What changes is who decides to switch them on, and when. Until now, a person made that call: turn on the heat when cold, move the lumbar support when the back hurts. Going forward, the car reads the passenger's state first and picks the right function itself. The point isn't how well any single feature performs — it's whether the car can bring several of them together at the right moment.
There's a clear limit on the health side, though.
"This doesn't mean the seat replaces a hospital and delivers a diagnosis. It means continuously observing multiple vital signs and the passenger's condition to catch potential risks early and support an appropriate response."
A car cabin isn't a clinical setting. There's vibration from the road, people keep moving, and clothing thickness and posture vary from one person to the next. Being able to read heart rate and breathing is one thing; using that data to diagnose an illness is another. The realistic job here is noticing change, not diagnosing disease — watching for signals that stray from someone's baseline, then alerting the passenger or routing the response through other vehicle functions. That still requires proving out sensor accuracy and cutting down false alarms.




The seat gathers data (sensing), reads the state (deciding), and carries out an adjustment (acting) — a three-step loop, and the basic framework behind XPENG's idea of an "intelligent seat."



From Memory to Prediction

If a seat can read the body's current state, the next step is anticipating what it will need.
That means combining weather, time of day, route, expected drive length, and a user's past habits. On a cold winter morning, it can warm the seat before someone even gets in. Ahead of a long drive, instead of waiting for fatigue to build and then switching on massage, it can adjust posture and pressure distribution beforehand. If someone sits differently than usual, the seat doesn't just replay a saved position — it can reset the backrest, cushion, headrest, and leg rest to match.
This is where it departs from a standard memory seat. A memory seat saves the position a person once found most comfortable and calls it back up. But comfort isn't fixed — a thirty-minute commute calls for a different posture than three hours on the highway, and a tired day is different from a rested one.
Personalization is shifting from remembering one position to finding whatever position fits the moment. It also matters when a user overrides an automatic adjustment: if the seat raises the lumbar support and the person lowers it again, the system can choose differently under the same conditions next time. Repeat that enough, and the number of times someone has to adjust the seat by hand drops on its own.
None of this needs to stay tied to one car, either. Seat position, support level, temperature, massage style, and usage habits can sync to a user account and carry over to a different vehicle. That works whether someone switches cars, rents one, or gets into a shared vehicle — the seat identifies the user and reconfigures itself. For something like robotaxi service, where the car changes every ride, this matters even more.
Today's memory seat means the car remembers the driver. Going forward, the setting won't just follow the car — it can follow the person. And that's where data questions start. Seat position and temperature are simple enough to store, but once body shape, weight, posture, heart rate, and breathing enter the picture, the stakes change. There needs to be a standard for what gets kept, what gets sent to another vehicle, and whether a user can review or delete it. Deciding who — the OEM, the mobility platform, or the seat supplier — manages and answers for that data is a problem no one can avoid once this reaches real service.
Ma Jia's presentation pointed to personalization traveling beyond a single car, but it didn't get into specifics on this front. The more a seat understands a person, the more personal the information it handles becomes — a question worth weighing apart from how convenient the technology is.




Predictive adjustment that learns from weather, routes, and habits to prepare a seat state in advance — and personalization that follows a user from one car to another through an account.



Level 3 Keeps You Awake. Level 4 Lets You Sleep

As autonomous driving moves up the scale, what a seat needs to do changes with it.
At Levels 2 and 3, the system handles part of the driving, but the driver stays tied to the task. At Level 3 specifically, the driver has to be ready to take back control the moment the system asks. What matters is checking gaze, posture, fatigue, and alertness, and making sure the driver can refocus when needed. A seat that knows the driver's physical state can play a role here too — correcting an overly relaxed posture, or using massage and vibration to raise alertness ahead of the handoff.
At Level 4, it flips. The driver becomes a passenger and no longer needs to face forward. Seats can slide along long rails or rotate so passengers face each other, and backrests and leg rests can reconfigure the cabin for rest or work.
The same intelligent seat has to keep a driver from dozing off at Level 3 and let a passenger relax at Level 4. That gap shows the seat can't operate independently of the car's driving state — it needs to know which autonomous function is active, when the driver has to take over, and whether the current seat position is even allowed under those conditions.
That puts the 128-degree zero-gravity pose in a different light. Reclining the backrest that far, or rotating the seat, helps with comfort but creates a much harder problem in a crash. Change a passenger's position and posture, and the way a seatbelt restrains the body changes with it — so does the relative position of the airbag. The path the body follows in a collision no longer matches a standard seated posture.
So as a seat gains more freedom of movement, the seatbelt, airbag, and occupant-detection and restraint systems all have to change alongside it. Building a comfortable position and protecting a person safely in that same position are two different engineering problems.




At Levels 2-3, the seat keeps the driver alert; at Level 4 and beyond, it supports a resting passenger. The same seat takes on opposite roles depending on the level of autonomy.



What Matters More Than Reclining to 128 Degrees

Future seats can carry even more functions. Directional speakers and active noise cancellation in the headrest could give each seat its own sound space, so one passenger watches a movie while another listens to music or takes a call. Haptic actuators could translate video or game motion into touch, and lighting, smart glass, and climate control could combine to create a different environment for each person in the same cabin.
But the number of features isn't the point. What Ma Jia describes as the future seat comes down to moving the functions that already exist in step with a passenger's state. He described it as "an emotional companion that understands the user in advance and offers the right service first — and a partner that watches over their health."
A backrest that reclines to 128 degrees is the easiest way to see this shift. But the angle itself doesn't make a seat intelligent. What matters more is knowing who needs that position, when it's safe to move into it, and how to keep someone comfortable and safe once they're there.
The next competition in seating begins beyond 128 degrees.

AEM(오토모티브일렉트로닉스매거진)

---