Intelligent driving technology is undergoing leapfrog development from highway scenarios to complex urban road scenarios. This progress entails not only the evolution of algorithmic models from perception to cognition but also a profound transformation in the layout of underlying perception hardware. Early autonomous driving solutions focused entirely on the area directly in front of the vehicle, adopting a single long-range LiDAR (Light Detection and Ranging) mounted on the roof or grille to detect obstacles 200 meters or farther ahead, ensuring longitudinal safety during high-speed driving.
With the mass production and deployment of urban Navigate on Autopilot (NOA) features, the scenarios that vehicles need to handle have become extremely complex. Frequent lane changes and cut-ins, unprotected left turns, and pedestrian avoidance on narrow streets have exposed the limitations of the original forward-facing perception with a simplistic field of view. For this reason, lateral LiDARs have been widely adopted.
Lateral Perception Imperative Driven by Urban NOA
In relatively closed scenarios with standardized traffic flow rules such as highways, core driving tasks mainly include lane keeping, distance control, and timely lane changes. In such cases, the sensor detection range is the top priority metric; as long as the system can perceive far enough in advance, it can secure sufficient reaction time.
However, once vehicles enter bustling urban streets, road conditions change qualitatively. The core task of lateral perception is no longer limited to monitoring rear-approaching vehicles in adjacent lanes, but focuses more on the precise detection of tiny, irregular, and dynamically changing near-field obstacles.
Although vision systems boast inherent advantages in semantic recognition, they suffer from shortcomings in spatial positioning accuracy and environmental robustness. Modern smart cars are equipped with high-definition cameras that build a panoramic perception space via multi-camera stitching technology. Nevertheless, as passive sensors, cameras experience a sharp decline in perception confidence when exposed to intense direct sunlight, abrupt light transitions at tunnel exits, or extremely low-light night environments.
The active emission characteristics of LiDAR allow it to remain unaffected by light fluctuations and deliver centimeter-level distance data, which is critical for addressing the common scenario of sudden hidden obstacle emergence in urban environments. Lateral LiDARs effectively compensate for the short-range ranging errors of cameras, providing more reliable physical spatial support for the vehicle decision-making system.
Furthermore, the integration of lateral LiDARs greatly alleviates blind spot risks during turns at complex intersections. When performing unprotected left turns or passing through large intersections, front-mounted LiDARs are constrained by a limited field of view (FOV) and fail to cover non-motor vehicles and low-height obstacles approaching from the vehicle sides.
In contrast, laterally mounted LiDARs can detect and continuously track these targets at an earlier stage, enabling accurate trajectory prediction. This capability is particularly vital for coping with mixed traffic scenarios involving non-motor vehicles and pedestrians. The robust performance of urban NOA systems largely relies on this near-field protective barrier built by lateral LiDARs.
From the perspective of safety redundancy logic, advanced autonomous driving requires the system to undertake greater operational responsibilities, especially after entering the conditional automation stage. Failure or misjudgment of a single type of sensor could lead to severe consequences.
The 3D point cloud data captured by lateral LiDARs enables cross-validation with data from cameras and millimeter-wave radars. When rain and fog obscure road curbs from camera perception, LiDAR can clearly outline curb contours through differences in reflection intensity. When millimeter-wave radars filter out stationary objects and fail to detect illegally parked roadside vehicles, LiDAR can accurately identify their physical boundaries via 3D modeling. This multi-dimensional cross-validation mechanism is the key to preventing intelligent driving system shutdowns and false braking under extreme operating conditions.
Technical Trade-off Between Long-Range LiDAR and Short-Range Solid-State LiDAR
Current mass-production lateral LiDAR solutions mainly adopt two distinct technical approaches. The first follows the performance balance principle, deploying two long-range LiDARs with specifications consistent with the front main LiDAR on vehicle sides (usually on fenders or both sides of the bumper). These LiDARs adopt a hybrid solid-state architecture with built-in precision rotating mirrors or galvanometer scanning systems.
The biggest advantage of this solution lies in unified perception performance. With a detection range of over 200 meters, these lateral LiDARs deliver clear short-range perception and enable early detection of fast-approaching side and rear vehicles during high-speed lane changes, providing the system with ample response margin. For automakers pursuing full-scenario perception capabilities, this solution generates abundant high-line-count point cloud data, helping improve algorithm recognition accuracy.
However, long-range LiDARs come with substantial cost pressure. High-performance scanning LiDARs maintain a high unit price, and the addition of two extra units significantly increases vehicle-level hardware costs. Additionally, such LiDARs typically feature a narrow vertical FOV of approximately 25 degrees. A high mounting position may create undetectable blind zones beneath the vehicle sides. Moreover, their sophisticated mechanical scanning structures, when laterally installed close to vehicle edges, are more vulnerable to road bumps and mud splashes, imposing higher requirements on sensor protection and cleaning systems.
The second approach follows the targeted blind spot filling principle, adopting short-range solid-state LiDARs specially optimized for near-field perception. Most of these LiDARs operate on the Time-of-Flight (ToF) principle and adopt array emission technology with no internal mechanical moving parts, hence the name pure solid-state LiDARs.
The core advantage of this solution is its superior FOV performance. Some advanced blind spot filling LiDARs achieve a vertical FOV of 70 degrees or even 90 degrees, covering targets from road lane lines to high roadside objects. Although their detection range only spans 30 to 50 meters, this level of detection accuracy and coverage fully meets the operational requirements of intelligent driving systems in low-speed scenarios, including complex intersection turns, narrow road traversal, and automatic parking.
Featuring a simple structure, pure solid-state blind spot filling LiDARs are compact enough to be discreetly embedded on vehicle sides. They outperform mechanical scanning LiDARs in reliability and boast a significant cost advantage.
With advancing technological maturity, the unit price of such blind spot filling LiDARs has dropped to 2,000 to 3,000 RMB. For mainstream automotive manufacturers aiming for large-scale popularization of intelligent driving functions, this solution that balances performance and cost is highly competitive.
Comparative analysis shows that long-range LiDARs and short-range solid-state LiDARs are not simply substitutive but serve different functional objectives. Long-range LiDARs act as all-round detectors for early long-distance potential threat identification, while short-range solid-state LiDARs function as wide-angle monitors to eliminate every near-field blind spot in complex driving scenarios.
Future Perception Evolution and Mainstream Configuration Forecast
As intelligent driving perception solutions continue to iterate, the industry has reached a convergent consensus on lateral LiDAR configuration. Following the single-LiDAR exploration period in previous years, intelligent driving hardware layout has fully entered a new era of multi-LiDAR fusion since 2024.
Judging from current technological evolution trends, the combined configuration of “1 long-range main LiDAR + 2 short-range blind spot filling LiDARs” is highly likely to become the mainstream solution for advanced intelligent driving in the coming years. This configuration realizes full-vehicle surrounding perception coverage at the physical level and achieves an optimal balance between cost and performance.
In this three-LiDAR system, the long-range LiDAR mounted on the roof or front center is responsible for longitudinal long-range perception during high-speed driving, monitoring static obstacles and fast-moving targets beyond 200 meters. The symmetrically installed short-range blind spot filling LiDARs on both vehicle sides leverage their ultra-wide vertical FOV to monitor near-field lateral environments.
This configuration enables vehicles traveling in complex urban areas to clearly perceive cutting-in electric two-wheelers from the side and rear, as well as low road curbs merely a dozen centimeters high. This long-and-short range collaborative strategy effectively reduces system false braking rates in complex environments and improves overall driving smoothness and ride comfort.
Pure solid-state technology will serve as the core driving force for the future iteration of lateral LiDARs. Improved chip integration will further accelerate the cost reduction of LiDAR products. Early LiDARs relied on expensive discrete optoelectronic devices, while the future development trend focuses on integrating emission, reception, and signal processing functions into a small number of silicon-based chips.
This innovation will not only reduce sensor size and power consumption but also transform LiDARs from high-precision optical instruments into standardized, mass-producible electronic components. In the future, even family cars in the 150,000 RMB price bracket are expected to be equipped with comprehensive lateral blind spot perception solutions, realizing the popularization of high-quality intelligent driving technology.
With the rollout of L3 autonomous driving pilot access policies and the clarification of legal liability for accident identification, the popularization of lateral perception hardware will be further accelerated. L3 and above intelligent driving systems require stable vehicle control capabilities even in extreme environments or in the event of single-sensor failure. Lateral LiDARs not only enhance perception capability but also constitute an indispensable part of safety redundancy. It is predictable that future lateral perception technology will evolve beyond basic object recognition and achieve in-depth integration with vehicle dynamic control systems.
Concluding Remarks
The adoption of lateral LiDARs marks the maturity of autonomous driving technology. It compensates for the inherent blind spots of visual perception, addresses perception challenges at complex urban intersections, and provides essential safety redundancy for advanced intelligent driving systems. In the technical selection between long-range scanning LiDARs and short-range pure solid-state LiDARs, the industry is converging toward a composite solution featuring targeted blind spot filling and complementary long-and-short range perception. This configuration guarantees comprehensive driving safety while ensuring controllable mass-production costs.
Driven by the popularization of chip-based technology and supply chain scale effects, LiDARs are shedding the label of high-cost sensors and becoming standard safety equipment for intelligent vehicles. Future autonomous driving systems will build an all-weather, omnidirectional, high-precision digital safety protection network through distributed LiDAR perception across the vehicle body.