Pay attention to and often read the tram lab punch card to think about new energy insights
Hardware determines the cap, software determines the experience.
Written by: CH_ Chen Han
After several months of thinking and preparation, at the end of 2021, we formulated the "Tram Lab Assisted Driving Evaluation Standard - 2022 Edition", and used this new set of standards to conduct a large evaluation of Tesla Model 3, NIO ES6 and Ideal ONE for assisted driving.
The differences in assisted driving hardware of these three cars are obvious:
Tesla Model 3 is equipped with two self-developed FSD autopilot chips, with a final force of 144TOPS;
Ideal ONE is equipped with two Horizon Journey 3 autopilot chips, with a final strength of 10TOPS;
Weilai ES6 is equipped with a Mobileye EyeQ4 autopilot chip, and the computing power is only 2.5TOPS.
However, our evaluation standards do not include a direct evaluation of hardware, because no matter how strong the hardware is, it also needs to be reflected through software and functional experience.
So, what is the assisted driving experience of these three cars? The stronger the computing power, the better the experience?
Click on the video to see the answer
Not convenient to watch videos? That's okay, the provincial little assistant has drawn the focus for you:
1. Test the route
2. Test data
3. Interaction design evaluation
Tesla Model 3:
To start assisted driving, you only need to quickly flick down the right lever twice. However, this version of the NOA hardly automatically regulates speed at high speeds, often requiring the driver to manually regulate the speed. Even if the choice of lane change does not require confirmation, it is still necessary to turn the steering wheel to confirm each lane change, and the force of this rotation is very difficult to grasp. In addition, it has no voice reminder, and it takes time to familiarize yourself with what the reminder message corresponds to different warning tones, which is very unfriendly to novices.
NIO ES6:
The assisted driving system supports one-button opening, automatic speed regulation, automatic lane change without confirmation, and the speed of opening the navigation assisted driving is the fastest, and the automatic recovery speed after the system exits is also the fastest. At the same time, in terms of information reminders, in addition to the reminders of vehicles and meters, there are also NOMI playful voice reminders, and the safety risk grading warnings are also relatively clear.
Ideal ONE:
Assisted driving is to flick the right lever inward twice in a row. It has a more exquisite assisted driving simulation interface, which can restore more road information on the car machine, and also has a rich information reminder such as voice, in addition, it also has a unique grip force sensing steering wheel in 3 cars, avoiding the problem that the other two cars need to turn the steering wheel to confirm. However, due to the frequent exit of navigation assisted driving, it has been relegated to lane keeping in front of the ramp many times, and it will also be downgraded before passing through the tunnel, so the convenience points have been lost.
4. Evaluation of single lane capability
All three cars can achieve timely and gentle acceleration and deceleration control, the cornering ability is also at an excellent level, and the ramp sharp bends in the test route can be easily handled, but in terms of details, the Model 3 cornering ability is still the strongest and most stable, and the Model 3 has a significant ability to cope with sudden insertion into the vehicle.
5. Multi-lane capability evaluation
Model 3's lane change ability is obviously above the other two cars, basically only need to have 2 parking spaces, Tesla Model 3 can successfully complete the lane change, its lane change choice is more decisive, lane change action is also quite "old driver".
In this test, we were also pleasantly surprised to find that model 3 has a certain "game" ability when changing lanes, when it changes lanes in a small space, it will first probe out a part of the body, and then judge the side rear vehicle to let go, if the side rear vehicle is obviously decelerated, it will then perform lane change action, if the distance is getting smaller and smaller, it will choose to return to the lane.
After many OTAs, NIO ES6's NOP change ability and lane change efficiency have been greatly improved, and in the case of a grasp of the situation on the other side, it can also do the light lane change in one go, but the change of lane is not as crisp as Model 3, and the required lane change space is significantly larger than Model 3.
The ideal ONE is the most Buddha of the 3 cars, and the least willing to change lanes. Basically, it can not change, it will not change, and the action of changing lanes is also the most moderate, but it will not appear to be very procrastinating.
6. Ramp ability evaluation
This test 4 ramp Model 3 passed 2 complete. Tesla's NOP is not available on many urban expressways or loops, and because it relies more on visual recognition, noops are withdrawn after encountering a little rain in this test. And the Model 3's performance in terms of ramp traffic efficiency is really bad, because it strictly limits the speed of 40km/h in the ramp, and it is impossible to manually adjust the maximum speed limit, resulting in a lot of psychological pressure every time we enter the ramp.
NOP can be used on highways and urban expressways covered by Baidu high-precision maps, with wide coverage and stable use. It completed 3 of the 4 ramps in this test, and the success rate of ramp traffic is high. And on the ramp, the ES6 supports manually adjusting the speed limit, but because of the limitation of the cornering ability, the speed in the corner is still a little slower than that of human driving.
The ideal ONE was affected by the frequent downgrade issue mentioned earlier, with only 1 of the 4 ramps completed.
After the test, we also communicated with the ideal autonomous driving product manager, who said that this situation was mainly caused by the lack or mismatch of high-precision map data, and for safety reasons, the system would choose to downgrade. They will refine the data as soon as possible and optimize the reminder of system degradation. Reduce the degradation rate and improve the user experience.
7. Conclusion
Combined 4 scores:
NIO ES6 76 points, ranked first
Tesla Model 3 73 points, ranking second
Ideal ONE 70 points, third place
Although the autopilot chip has the lowest computing power, with the continuous optimization of the software, NIO's autonomous driving engineers have developed a more complete experience of the driver assistance system, which once again proves the importance of software and tuning for assisted driving.
Tesla has once again proved to us the unfathomable potential of pure vision solutions and Tesla's ability to lead the way in the field of autonomous driving. FSD, Full Self-Driving is Tesla's great vision to move forward firmly, not the publicity of the new bee. It's just that, as always, less humane care, and lack temperature to use.
The first version of navigation assisted driving has been remarkable, and it should have a lot of room for improvement through continuous optimization through OTA. More importantly, the ideal ONE's navigation assisted driving software is standard for all models in 2021, while the software of Tesla NOA and NIO NOP is priced at more than 30,000 yuan. In comparison, this free navigation assistance driving system has such a performance, which is obviously really fragrant.
In general, after understanding the boundaries of their respective capabilities, under the premise of correct use, the auxiliary driving systems of these three cars are practical and easy to use, which can effectively reduce the fatigue of drivers driving long distances and improve the safety of travel to a certain extent.
Please note that this is "under the premise of correct use", what is correct use? For example, we now know that Tesla is driving slowly on the ramp, so after entering the ramp, we need to take over the driving and take the initiative to speed up.
More importantly, whenever and wherever you use assisted driving functions, you must always pay attention to the road conditions and be ready to take over the vehicle.
We should not be overly mythical and over-reliant on assisted driving systems, and we should not treat assisted driving as automatic driving. At the same time, we should not "choke on food" and completely deny the positive role and development results of assisted driving.
Today's topic
Which car would you like to see in the Assisted Driving Review?
Feel free to leave a message in the comment area
Scan the code to add the operation assistant and make notes
Join the Tram Lab AC group