Standard LiDAR Across All Models: Tech Democratization or Cost Illusion?

LiDAR went from a high-end optional extra to standard equipment across all trims in just three years. But what does Deepal S05's "standard on all trims" actually mean? It's worth breaking down.

The Deepal S05 is positioned as the "Global Fashion LiDAR Smart SUV," with pre-sales starting tomorrow. The suspense isn't whether it has LiDAR, but: who makes this radar, how many lines does it have, what scanning method does it use, what is its ranging accuracy, and—can it truly run city NOA?

First, let's look at the industry status quo. In 2024, mainstream LiDAR solutions have diverged into three paths:

• Semi-solid-state rotating mirror (Hesai AT128, RoboSense M1): 128 lines, detection range 200m+, cost approx. 1500-2000 RMB
• Pure solid-state Flash (Lokar, Huawei): No mechanical parts, but short detection range (<100m), cost can drop below 1000 RMB
• Mechanical rotating (older models): Has basically exited the passenger vehicle OEM market

If the Deepal S05 comes with LiDAR standard on all trims, it means even the base model must include one LiDAR unit. Given current supply chain costs, the BOM cost for a 128-line semi-solid-state LiDAR is roughly 1800 RMB. The Deepal S05 is expected to be priced between 150k-200k RMB. For cars in this price segment to pile on smart driving hardware, the usual choices are: either use a pure vision solution, or use one Flash radar for perception redundancy.

But Deepal's "standard on all trims" might have some hidden tricks. I checked Deepal's previous S7 smart driving solution; it used Huawei MDC 610 + 5R11V (5 millimeter-wave radars + 11 cameras), with no LiDAR. Suddenly adding it now, and emphasizing "standard on all trims," suggests it's likely not self-developed, but rather a packaged solution from a Tier 1 supplier.

[!note]

Standard LiDAR on all trims does not mean standard city smart driving on all trims. Currently, LiDAR mostly serves as "perception redundancy," providing safety backup for basic functions like AEB and LCC, rather than being a necessary condition for advanced navigation assistance.

Let's look at a typical engineering implementation comparison:

Solution A: Pure Vision + Millimeter-Wave Radar
├─ Cost: ~800 RMB
├─ Dependency: High-compute chips + massive training data
└─ Risk: Performance degradation in extreme weather

Solution B: Vision + 1 Solid-State LiDAR
├─ Cost: ~2200 RMB (including radar + algorithm adaptation)
├─ Dependency: V2X fused perception, slightly lower compute requirements
└─ Advantage: Higher robustness in night, strong light, rain, and fog

Deepal chose Solution B, but that doesn't mean it possesses full-scenario smart driving capabilities. The key lies in: whether this LiDAR's "effective pixels" and "frame rate" support real-time mapping and localization. If it's just a 16-line low-frame-rate radar used to trigger AEB in heavy fog, its value is far less sexy than advertised.

I guess the Deepal S05's LiDAR configuration is as follows:

• Supplier: RoboSense (widely adopted by BYD and Xpeng, large production capacity, cost advantage)
• Model: M1 or M2 semi-solid-state, 128 or 96 lines
• Interface: Ethernet output of LiDAR point cloud, performing early fusion with cameras within the domain controller
• Function: Primarily used for static obstacle detection on city roads, lane line assistance (outputting boundaries when lane lines are blurry), and following target recognition in congested traffic

But the problem is, "standard LiDAR on all trims" will eat into the vehicle's profit margin. Calculating based on a starting price of 150k RMB, the radar cost alone accounts for 1.2%, plus the adapted domain controller (at least an additional 1000 RMB), the total smart driving system cost share could exceed 8%. This is quite aggressive for a car in the 150k RMB class.

Unless Deepal's LiDAR is a "downgraded version"—for example, 32 lines or 16 lines, covering only a 120-degree forward FOV with an 80m detection range. Such radars can be pressed down to under 500 RMB in cost, but they can only support L2-level auxiliary functions like automatic emergency braking and forward collision warning, completely unable to support city NOA.

[!tip]

How to judge the true value of a car's LiDAR: See if it supports features like "memory parking," "city navigation," and "valet parking." If it only has basic AEB, then this radar is likely low-cost tech stockpiling, not a genuine perception upgrade.

Finally, back to pricing strategy. Deepal S05's "standard on all trims" feels more like a marketing positioning move. In the 150k-200k RMB range, no other car currently dares to make LiDAR standard on all trims. Once they do, it establishes a cognitive anchor of "tech safety" in consumers' minds. Even if actual functionality is limited, it can suppress rivals like BYD Song and Geely Galaxy in public opinion.

But from an engineering perspective, the challenge of standard LiDAR on all trims lies in the scale effect of algorithm adaptation. If the base trim gets only one forward radar, while mid/high trims get three (forward + two sides), the underlying perception fusion algorithm must simultaneously support both 1-radar and 3-radar hardware topologies. This means the domain controller's software architecture must be sufficiently abstract, otherwise code maintenance costs will grow exponentially.

```

if (hard

Original link: https://www.ithome.com/0/983/026.htm