Fortsense Technologies claims its single‑chip RGBD spatial cameras will merge color imaging and depth sensing, potentially replacing separate camera‑LiDAR stacks in autonomous vehicles and robots. The company’s progress on SPAD‑SoC integration, omni‑directional scanning optics, and a roadmap to 4 MP automotive parts is real, but the technology still faces resolution, power, and safety‑grade certification hurdles before it can displace established sensor suites.
Claim → Reality → Caveats
Fortsense Technologies announced that its single‑chip RGBD spatial cameras will soon combine high‑resolution color imaging with direct‑time‑of‑flight (dToF) depth measurement on a single silicon die. The company positions the solution as a drop‑in replacement for the traditional camera + LiDAR stack used in autonomous driving and industrial robotics.
What’s actually new?
| Feature | Detail |
|---|---|
| SPAD‑SoC integration | Fortsense has been mass‑producing silicon‑photomultiplier (SPAD) arrays with on‑chip signal processing since Q4 2025. The chips are fabricated in a 28 nm CMOS process and include a custom analog front‑end for single‑photon detection. |
| Omni‑directional optical scanner | Their proprietary scanning module claims 80 % laser utilization by using a diffractive optical element that spreads the emitted pulse over a 360° field while preserving range. This enables solid‑state LiDAR performance out to 200 m with a modest 5 W laser budget. |
| Resolution roadmap | Current production parts deliver 300‑400 k pixel depth maps (≈640 × 480) for robotics. A 4 MP RGB sensor paired with a 1 MP depth array is slated for 2027, targeting automotive‑grade ADAS/AD requirements. |
| Manufacturing pedigree | Fortsense is the only Chinese firm that controls the full SPAD stack—from wafer design to automotive‑qualified packaging—allowing them to claim a cost advantage over imported LiDAR modules. |
The company has already shipped tens of thousands of chips to customers such as the Zeekr 9X electric SUV, where the sensors are used for low‑speed parking assistance and short‑range obstacle detection.
Why it matters
Physical AI systems (autonomous cars, warehouse robots, drones) need accurate 3‑D perception to make safety‑critical decisions. Traditional monocular or stereo cameras infer depth from disparity, which can be brittle under low‑light or texture‑poor conditions. Adding a dedicated depth sensor mitigates those failure modes, but it also adds cost, weight, and calibration complexity.
If a single chip can deliver color + depth with comparable range and reliability, system integrators could simplify hardware design, reduce bill of materials, and potentially lower the overall cost of perception stacks. That would be a tangible engineering win, even if the market impact is gradual.
Limitations that remain
- Pixel density vs. depth accuracy
- The current 300‑400 k pixel depth maps are adequate for indoor robotics but fall short of the high‑resolution point clouds produced by state‑of‑the‑art LiDARs (often >1 M points). For high‑speed highway driving, sparse depth can miss small obstacles.
- Power and thermal budget
- SPAD arrays operating at 5 W laser power plus on‑chip processing can generate >2 W of chip dissipation. In a compact automotive module, that heat must be managed to meet ISO 26262 functional safety standards.
- Safety certification timeline
- While Fortsense has automotive‑grade packaging, the ASIL‑D certification required for primary perception sensors is still pending. The roadmap to 2027 does not guarantee regulatory approval by then.
- Algorithmic integration
- Merging RGB and dToF data in real time demands sensor‑fusion pipelines that are still maturing. Existing perception stacks (e.g., Apollo, Autoware) assume separate data streams; rewriting them for a unified sensor adds development overhead.
- Competing approaches
- Other firms are pushing solid‑state LiDAR with MEMS mirrors and high‑resolution stereo cameras that already meet automotive standards. Fortsense’s cost advantage hinges on the claimed 80 % laser utilization, but independent benchmarks are not yet public.
Bottom line
Fortsense’s single‑chip RGBD cameras represent a real engineering progression: integrating SPAD‑based depth sensing with a conventional image sensor on a single die, backed by an in‑house optical scanner. The announced roadmap—dual‑chip prototypes now, a fully integrated chip by 2027 with 4 MP RGB—shows a concrete development timeline.
However, the gap between prototype performance and the stringent requirements of highway‑level autonomy remains sizable. Until the company publishes independent benchmark data, demonstrates ASIL‑D compliance, and shows that the fused sensor can sustain high‑density point clouds at long range, the claim that RGBD will replace discrete camera‑LiDAR stacks should be treated as a long‑term possibility, not an imminent industry shift.
Image credit: 
Comments
Please log in or register to join the discussion