eastwest connect

NIO’s NOP+ performance status and test drive experience example


Source: Zhixing Jiadao (Smart Driving)

 

Recently, on March 7, Zhixing Jiadao (Smart Driving) introduced NIO’s NOP+ and their test drive experience, as presented below.

On February 20, NIO announced that the NOP+ Beta trial opened for users to register. We also experienced NIO’s NOP+ for the first time. Taking NIO’s NOP+ as an example, we will talk to you about the current situation of highway & ramp navigation-assisted driving, and answer some questions.

What are the advantages of NOP+?

The difference between NOP and NOP+ can be divided into hardware and software differences. The hardware uses 4 Nvidia Orin chips, with a computing power of up to 1016TOPS, standard 1550nm long-range lidar and so on. In terms of software, it has also fully switched from the previous partial self-research to full-stack self-research.

Therefore, the greatest significance of NOP+ actually symbolizes NIO’s first truly self-developed intelligent driving product.

Reflected in the specific road conditions, the core feature of NOP+ is that it can intelligently think and act dynamically according to road traffic conditions. In human terms, it drives more like a human, rather than a rigid and cold machine.

Driving experience

How about the specifics, you will know when you drive on the road. Our NOP+ experience this time starts from Huangpu District, Guangzhou to Conghua Bus Station. The whole journey is 61 kilometers, of which 51 kilometers can turn on the NOP+ function.

It is worth mentioning that to enable NOP+, you need to switch to Enhanced Pilot Assist in the assisted driving settings first, and the preparation work is completed after setting the navigation. The reason for setting the navigation is because this is also dependent on high-precision maps. For assisted driving, the high-precision map of NIO NOP+ is provided by Tencent. This is also the first time that Tencent’s high-precision map has been used in a car.

After setting up the navigation, you can see a blue dot in the navigation route, which means that the NOP+ function can be turned on from here, and the red dot is the end point.

After turning on NOP+, an obvious perception is that the enthusiasm for changing lanes is significantly higher. If you are driving in the slow lane after turning on, then NOP+ will most likely choose the optimal lane immediately after turning on, usually the fastest lane or the innermost lane, instead of foolishly going to the end in the same lane.

As for the strategy of changing lanes and overtaking, there are three in summary:

On an open road section, if you encounter a slow vehicle ahead, the system can accurately identify it and change lanes in advance;

If there is a car in the target lane, the lane change will be initiated immediately when the car overtakes us and there is enough space, instead of rigidly waiting for the car in the target lane to drive far away before changing;

If there is a lane change initiated at the same time as the vehicle behind, the system will not retreat immediately in a particularly conservative manner, but will accelerate immediately while changing lanes like a human driver;
If you change lanes with the vehicle in front, the system will not retreat, but change lanes without accelerating.

The timing of entering the ramp to initiate a lane change is also very flexible. Before entering the ramp, if you are in the leftmost lane, NOP+ will initiate the first lane change at about 2 km from the ramp, the second lane change at 1.5 km, and the third lane change at 1 km.

However, according to other users, the time to merge into the ramp is not fixed, but is related to the navigation information. If there is congestion, the system will initiate a lane change in advance, and start looking for opportunities to change lanes earlier, which will help improve the ramp into success rate. However, we did not encounter ramp congestion, so we did not have the opportunity to verify.

In terms of speed reduction on ramps, NOP+ will adjust the speed limit to 70 km/h after entering the ramp. If a large curvature curve is encountered later, the system will actively adjust the speed according to the curvature of the curve, so compared to those who strictly follow the speed limit, the vehicle speed control logic of NOP+ is very smooth throughout the ramp, and there is no sense of fast and slow passages.

This is also the advantage of NOP+’s intelligent speed limit adjustment. It does not follow the minimum speed limit very strictly, but adjusts the speed according to the specific situation, avoiding the embarrassing situation of being double-flashed and honked by the rear car. This function extends the personalized cruise speed function, even on the straight road, sometimes it will not strictly follow the speed limit, and will switch flexibly for the sake of traffic efficiency.

However, NOP still has some deficiencies. For example, when encountering congested road conditions, NOP+ brakes are too heavy, and sudden braking may cause rear-end collisions.

In addition, in congested road conditions, if there is a vehicle blocking the road, NOP+ will not predict the deceleration when the blocking vehicle starts to change lanes, and it will suddenly come to a stop when the blocking operation is about one-third of the way, which is easy to startle.

The following distance under low-speed road conditions is also too cautious. Basically, the distance from the car in front can still accommodate two cars. Although the following distance is adjustable, the adjustment method is not intuitive and convenient enough.

This also just explains why NOP+ is only applicable to highway & ramp sections, because there are too many congested urban road conditions, and there are still many places where NOP+ can be optimized in dealing with congested road conditions.

Autonomous driving, slow is fast

Judging from the above test results, the functions that NOP+ can achieve are not very subversive, and are not far behind other NOAs. Indeed, there is no difference in function between NOP+ and NOP, but a set of things has been changed from the underlying algorithm and hardware composition. Although the functions are similar, the experience has been upgraded a lot.

The + sign of NOP+ represents the possibility of the future.
The computing hardware used by NIO on the NT1.0 platform is the EyeQ4 chip provided by Mobileye. The cooperation mode of NT1.0 is that Mobileye masters the underlying hardware and perception algorithms of intelligent driving, and NIO can only carry out secondary development on top of this. Therefore, it can be said that the soul of NT1.0 is actually controlled by Mobileye.

This also means that after the NT2.0 technology platform adopts the more open Nvidia Orin chip, NIO’s NOP+ has truly realized full-stack self-development. Although the functions are close, it is of great significance to NIO.

On this basis, NIO can realize its own rapid iteration, which is the so-called growth. Harry Wong, head of NIO’s autonomous driving product experience, said that starting from the NT2.0 technology platform, NOP+ developed by NIO can basically be iterated for all assisted driving functions.

At present, NIO has realized high-level intelligent driving in highway scenes, but it will take some time to enter the city’s open roads.