XOS 5.4.x Experience Part #1

一般来说，这是一个强有力的更新，有期待已久的改进和优化。在更新之前，Xpeng还启用了本地英语支持，这是自3年前以来梦想的惊人进步，但最终看到它发生了。我几乎失去了欢乐的一年。感谢我的管家，他明智地跟进了这个问题，并实现了它。我对这位管家鹏管家-珠桡的重大力。

[Al驱动程序-XNGP：在所有场景中使用端到端的大型模型。 / Al driver - XNGP: Use end-to-end large model for all scenarios.]
经验反馈 / My Experience: I’m not sure I understand this. The system still reports NGP unavailability for certain routes. I’m glad to see the switch to using the large model to plan and decide. I hope this experience catches on. Since the update, the car has been veering to the rightmost lane and sometimes occupying two lanes. What happened to the guardrail logic? Please check and fix it.

[Al停车：效率和体验都大大提高了 /Al parking: Efficiency and experience are greatly improved]
经验反馈 / My Experience: Yes, there’s an improvement in maneuvering. Even though it’s minor, it’s evident now that parking efficiency is much better. However, there are a few observations about the logic or conditions for remote parking that need to be updated. For instance, if all doors are closed, and then any of the other doors are opened within moments before the remote parking starts, the system should be able to resume. Additionally, sometimes the remote parking fails to work. This situation suggests that the system relies on persistent connectivity via a mobile network, but I don’t believe this enhances the experience. While monitoring the car via the mobile network is important, for remote parking, the car should be able to make autonomous decisions without that hard tethering for specific instances like emergency stop. The local use of sensors should more than suffice ability to remote park without hard-teethering.

[Al保镖：直播视频：支持紧急视频功能，始终保持车辆安全 / Al bodyguard: Live Video: Supports emergency video capabilities to keep vehicles safe at all times.]
经验反馈 / My Experience : To enhance the user experience, I kindly request that you eliminate the need to manually enable this feature each time. It’s an unnecessary burden. Please update the feature to allow for a one-time setup, ensuring that it remains active as a configuration item without the need for constant self-reminder to activate it. Additionally, within the Power Management page, I expect to find a comprehensive configuration plan for auxiliary device usage, including AI bodyguard cameras, that these devices will automatically shut down when the battery threshold for auxiliary devices is set.

[背面图像：智能驾驶状态转弯图像可以激活，驾驶更安全。 /Rear Side Image: Intelligent Driving State Turning Image Can Be Activated, Safer Driving.]
经验反馈 / My Experience: This year, this is one of the greatest driver support features. Thank you for finally addressing my previous complaint and optimizing it. The new feature now fully complies with an experiential use, providing drivers with peace of mind and awareness of their surroundings. However, there are a few additional requirements that need to be implemented.
1. First, include a rear video feed option. This option would allow drivers to have a full rear view, not just the rear sides, either in the driver’s cockpit display or the driver’s auxiliary display. When a driver turns on the turn signal for left or right, they currently only see the side rear view, not the actual rear. One scenario where this would be beneficial is when a skilled driver is making a turn. They don’t just rely on their left and right side mirrors; they also use the rear view mirror to ensure nothing is approaching from behind. This provides a sense of security and complete awareness. Therefore, this option should also be available for each turn signal, allowing drivers to see both the side video and rear video to enhance their confidence and awareness. This feature should also be applied to the self-driving mode use of turn signals and the emergency pop-up of the views when there’s an intent to warn the driver of a car approaching them from behind faster than normal.
2. Second, when the driver turns on hazard lights while the car is in “D” mode, show the left, right, and rear video feeds. For drivers who want to have the rear video feed as a permanent driving aid, they should be able to have it as a permanent video overlay on the center console or in the driver’s cockpit screen as another window in the center console (e.g., Map mode, Smart drive mode, + Rear mirror mode).

[Al底盘-智能底盘：鹰眼智能调整底盘，使旅程更加舒适 / Al chassis - Smart chassis: Eagle Eye intelligently adjusts the chassis to makethejourney more comfortable]
经验反馈 / My Experience is that it’s a good start. However, I found that the algorithm is too slow to predict road conditions and also adjust the suspension reaction in real time to mitigate the impact of road conditions.

[个性化操作系统2.0 / Personalized Operating System 2.0]
经验反馈 / My Experience: The split-screen feature enhances the utilization of the screen’s real estate. However, I would like to have the option to display hover cards on one of the split screens without the need to completely change the split area. The current split-screen configuration is quite restrictive, and I believe incorporating hover cards would significantly improve the usability of the split-screen feature.

[优化-基于端到端模型能力，XNGP极大地改善了城市场景中变道、绕道和加速和减速的体验，使其更接近人类驾驶行为。 / Optimization - Based on the end-to-end model capabilities, XNGP greatly improves the experience of changing lanes and detours and acceleration and deceleration in urban scenes, making it closer to human driving behavior.]
经验反馈 / My Experience: There’s a significant improvement, and I agree that it enhances the overall experience. However, the decision-making process still lacks a closed-loop connection between perception and localization. You’ll notice the absence of human input, such as pedal acceleration, in certain decision trees, including the use of turn signals. As a result, you’ll observe the car signaling left but turning right, or vice versa. To address this issue, the closed-loop between the decision model and actual localization requires more real-time or tactile feedback and feedback. This needs to be thoroughly addressed in the near future to enhance confidence. It’s reassuring to see that the car’s behavior is becoming less robotic, yet there are still many inconsistencies in its actions, reminiscent of an elderly woman trying to drive. While I agree that it’s becoming more human-like, it’s not the type of human driver we need on our busy roads to emulate - a granny. I hope to see further improvements in lane change, lane detection and perception, the time taken to execute changes, and the efficiency of executing lane changes themselves.

[变道和绕行路径：变道和绕行路径更像老司机，同时，进一步提高绕行的决策能力和在拥堵场景中切断的能力 / Changing lanes and bypassing paths: Changing lanes and bypassing paths are more like old drivers, and at the same time, further improve the decision-making ability of bypassing and the ability to cut in congestion scenarios]
经验反馈 / My Experience: There’s improvement, but it’s not as smooth as the “old driver” as you hint. Perhaps smooth as the old-man driving, because it’s full of granny like behaviors. The ability to slide in and out of lanes is better, and i agree that this improvement is closer to natural than robotic, but managing lane occupancy still needs marked optimization to ensure that changing lanes is done in a way that clearly ensures firm confidence, and also clear lane occupancy, not wander around multiple lanes.

[速度控制：速度控制更拟人化，启动反应更灵敏，同时减少不必要的减速造成的不适感 / Speed Control: Speed control is more anthropomorphic, with a more responsive start, while reducing the discomfort caused by unnecessary deceleration]
经验反馈 / My Experience: The speed control system has been improved, but there’s still a delay between optimizing speed to match traffic conditions and modulating steering when the speed is adjusted to match the driver’s decision. We need to consider scenarios where the discomfort of driving too slowly prompts the driver to press the accelerator pedal. This information is crucial for determining the appropriate location and making decisions about using the turn signal. However, there’s a missing logic that connects these factors, which results in a poor response rate that negatively impacts the sound and smoothness of the drive. So as an example, the car will signal right too late while it is making a left turn movement, or vice versa. This confuses the car behind greatly.

[优化-改善XNGP使用的连续性，并支持用户在更改导航瞄准器、添加路线点等时更改导航路线，XNGP将不再降级为LCC。同时，它减少了因道路异常而导致的XNGP系统退化 /Optimization - Improve the continuity of XNGP use, and support users to change navigation routes when changing navigation sights, adding route points, etc., XNGP will no longer be degraded to LCC.At the same time, it reduces the system degradation of XNGP due to abnormal road]
经验反馈 / My Experience: Well done! This is an excellent optimization. The result is clearly evident, with a distinct night/day difference. I hope a new feature can be implemented, which would allow users to select the next rest stop and have the car retain a path that facilitates easy exit from rest stops.

#OTA建议# #G9 XmartOS 体验#