TLDR¶

• Core Features: Hesai CEO David Li cautions that society and regulation are not prepared for rapid, widespread deployment of L4–L5 self-driving vehicles.

• Main Advantages: Lidar advances can reduce accidents and improve perception, but risk tolerance, liability frameworks, and public trust remain unsettled.

• User Experience: Early autonomous deployments show promise in constrained domains, yet mixed public reception and edge-case failures highlight fragility.

• Considerations: Even a single fatality per year from one company’s autonomous fleet could trigger existential backlash, despite large safety gains.

• Purchase Recommendation: Invest selectively in autonomous tech supported by robust safety practices, redundant sensing, transparent reporting, and realistic deployment timelines.

Product Specifications & Ratings¶

Overall Rating: ⭐⭐⭐⭐⭐ (4.8/5.0)

Product Overview¶

Autonomous vehicles (AVs) promise to dramatically reduce road fatalities and reshape mobility, but the path to broad societal acceptance is fraught with technical, ethical, and regulatory complexities. In this context, lidar—a laser-based sensing technology—plays a central role. By providing high-resolution, three-dimensional mapping of the environment, lidar augments cameras and radar to deliver robust perception, especially in low-light, high-contrast, or adverse weather conditions. Hesai, one of the leading lidar manufacturers globally, has been a prominent supplier to OEMs and robotaxi developers seeking dependable sensing stacks.

David Li, CEO of Hesai, recently underscored a sobering reality: while autonomous systems can theoretically reduce the nearly one million annual road deaths worldwide, society is not yet ready for a rapid rollout of high-autonomy vehicles. His caution highlights a crucial tension. On one hand, AVs may outperform average human drivers over time, mitigating fatigue, distraction, and impairment. On the other hand, public tolerance for machine-caused fatalities is extraordinarily low. As Li noted, a single death attributed annually to a specific company’s autonomous vehicle could trigger formidable public and regulatory backlash—potentially existential for the business—regardless of comparative safety gains.

First impressions of this position are compelling because they balance optimism with prudence. Li’s argument does not diminish the promise of AVs or the importance of lidar; rather, it calls for thoughtful deployment strategies that align technological capabilities with societal readiness. This approach emphasizes transparent metrics, clear liability frameworks, strong incident investigative processes, and responsible scaling.

To understand the stakes, consider the broader AV ecosystem: escalating investments, aggressive timelines, and increasingly complex urban pilot programs. Lidar-equipped vehicles are beginning to demonstrate reliable performance under well-defined constraints, such as geo-fenced routes and controlled speeds. Yet, edge cases—obscured pedestrians, unusual vehicle behavior, construction zones, unpredictable cyclists—continue to stress even sophisticated sensor fusion stacks. The public’s intuitive response to rare but high-profile incidents is a critical strategic variable. Companies must therefore manage expectations, communicate risk honestly, and cultivate trust through consistent safety outcomes and verifiable data.

In sum, Hesai’s perspective reflects the maturation of the AV industry. The technology is advancing quickly, but durable success depends on societal acceptance, clear regulation, and resilient business models that can withstand scrutiny when inevitable incidents occur. Lidar remains a cornerstone of this journey, enabling perception that is precise, repeatable, and transparent. The next stage will be defined not only by breakthroughs in sensors and AI, but by the quality of governance that frames their use.

In-Depth Review¶

A comprehensive review of today’s autonomous vehicle landscape must begin with the sensing stack where lidar provides a decisive advantage. Lidar systems emit laser pulses to measure distances, building dense point clouds that capture object shapes, positions, and motion. This 3D understanding complements camera-based semantic recognition and radar’s robustness to weather, forming a triad of perception that can be fused for reliable situational awareness.

Technical characteristics:
– Range and resolution: Modern automotive lidar can detect objects at long distances with fine angular resolution, enabling early detection and classification. High channel counts and refined scanning patterns deliver granular point clouds that support precise localization and obstacle recognition.
– Weather and lighting resilience: Unlike cameras, lidar is less affected by low light and glare; unlike radar, it can provide detailed geometry, improving detection of small or partially occluded objects.
– Real-time processing: Lidar point clouds demand significant compute for clustering, segmentation, and tracking. Efficient on-vehicle processing pipelines, optimized neural networks, and hardware acceleration are integral for timely decision-making.

From a performance testing perspective, AV developers typically evaluate lidar-enhanced perception across varied conditions: daytime and nighttime, rain and fog, urban density and open highways, work zones, and interactions with pedestrians and cyclists. Reliable performance comes from robust sensor fusion, redundancy (multiple lidar units or overlapping fields of view), and fail-operational designs that maintain situational awareness even when one sensor degrades.

Yet, the cutting edge in perception is only part of the equation. The AV stack includes:
– Localization: Combining lidar with HD maps and inertial/GNSS data to determine vehicle position with sub-meter accuracy.
– Prediction: Anticipating trajectories of surrounding agents—cars, buses, scooters, pedestrians—based on behavior models and learned patterns.
– Planning and control: Generating safe, efficient trajectories that adhere to traffic laws and social driving norms, then executing them through precise control of steering, throttle, and braking.

In early pilot programs, lidar-equipped vehicles tend to perform well within geo-fenced areas, benefiting from prior mapping and known operational design domains (ODDs). However, performance variability arises in unfamiliar situations: sudden road closures, non-standard signage, erratic human behavior, and unstructured environments. Mitigating these weaknesses demands continuous data collection, model retraining, and robust scenario testing.

*圖片來源：Unsplash*

David Li’s warning about societal readiness hinges on how we evaluate risk and accountability. For human drivers, society tacitly accepts a tragic baseline of global fatalities approaching one million per year. For machines, tolerance is far lower because responsibility is concentrated. When a single company’s vehicle causes a fatality, the accountability chain is clearer, and public perception becomes a decisive force. This asymmetry pressures AV companies to aim not merely for “better than average human” performance, but for safety margins that dramatically exceed expectations, along with transparent reporting and rapid corrective actions.

Regulatory landscapes vary by region. Some jurisdictions enable limited driverless operation under strict permits, incident reporting, and geo-fencing. Others adopt a cautious stance, requiring human safety operators or constraining routes. Uniform standards for metrics—disengagement rates, incident severity, safe driving criteria—are still evolving. An objective review suggests that resilient AV deployment strategies should include:
– Redundant sensing: Multiple lidars, cameras, and radar, cross-checking results to reduce single-point failures.
– Conservative planning: Risk-aware maneuvers that err on safety, even at the cost of efficiency.
– Incident automation: Immediate detection, secure logging, remote diagnostics, and transparent reporting to authorities and the public.
– Continuous validation: Large-scale simulation and replay of real-world edge cases, with measurable improvements over time.

The business implications are just as critical. A single high-profile incident can trigger pauses, audits, and reputational damage. Therefore, AV companies and their suppliers must invest in safety culture, third-party assessments, and clear communication strategies. Lidar manufacturers like Hesai play a pivotal role: they supply the perception backbone, collaborate on testing regimes, and contribute to system-level reliability alongside OEMs and robotaxi operators.

Ultimately, performance excellence must be paired with societal acceptance. That means setting realistic expectations about timelines, capabilities, and limitations, while demonstrating, with evidence, that AVs reduce risk meaningfully. Li’s statement reflects a pragmatic stance: proceed, but with the humility to recognize that technology alone cannot solve trust, governance, and ethics. Those domains require consistent leadership, and a willingness to slow rollout when safety data demands it.

Real-World Experience¶

Pilot deployments of autonomous vehicles in major cities offer a window into the user experience and societal reception. Riders often report smooth, consistent driving that avoids aggressive maneuvers common among human drivers. Vehicles equipped with lidar demonstrate strong lane keeping, predictable braking, and cautious turns, especially when navigating complex intersections or busy downtown corridors. In good conditions, the ride quality is reassuring.

However, real-world experience reveals nuances:
– Edge-case behavior: When encountering unexpected scenarios—say, a pedestrian darting from behind a parked van—lidar’s ability to capture geometry can help detect movement earlier than cameras under glare or low-light. Still, decision-making must balance false positives (overly cautious stops) with responsive avoidance. Excessive conservatism can cause traffic flow issues and frustrate nearby drivers, while insufficient caution risks accidents.
– Interaction with non-compliant actors: Human road users often deviate from norms—rolling stops, mid-block crossings, last-second lane changes. AVs tend to follow rules strictly, which can cause awkward interactions. A carefully tuned social driving model is essential to remain safe without becoming obstructive.
– Weather and infrastructure variability: Heavy rain, fog, or dirty sensors can degrade perception. Regular calibration and sensor cleaning regimes are vital. Construction zones, temporary signage, and unmarked detours further test the system’s adaptability.
– Incident transparency: When minor collisions or close calls occur, the quality of incident handling profoundly shapes trust. Clear logs, accountable explanations, rapid software fixes, and open communication with regulators and the public are indispensable.

User trust grows with repeated positive experiences and the perception that the system behaves consistently and predictably. Early adopters—tech-savvy riders and fleet operators—often appreciate the safety discipline and data-driven improvement cycles. Yet broader public acceptance depends on how companies navigate rare but severe incidents. As Li observed, even one fatality per year tied to a single company’s fleet could provoke outsized backlash, overshadowing aggregate safety benefits.

Fleet operations teams report that lidar-rich sensor suites simplify certain tasks: accurate mapping updates, better detection of small objects (debris, low-height obstacles), and more reliable cut-in detection on multilane roads. Maintenance practices—sensor cleaning, alignment checks, firmware updates—are routine but must be rigorous. Over-the-air updates enable rapid performance improvements, though change management requires careful validation to avoid introducing regressions.

An often-overlooked dimension is rider communication. Clear in-cabin status indicators, route explanations, and contingency messaging (e.g., why the vehicle stopped unexpectedly) reduce anxiety. Customer support that can contextualize incidents and answer safety questions further builds credibility. Companies that treat communication as part of the product experience fare better in public perception.

In mixed-traffic environments, lidar contributes to a calmer driving style. Vehicles can maintain adequate following distances and respond smoothly to dynamic conditions. Over time, consistent behavior reduces the likelihood of collision cascades caused by abrupt maneuvers. Still, AVs must adapt to local driving cultures and infrastructure quirks—what works in one city may not translate seamlessly to another.

The real-world takeaway aligns with Li’s caution: AVs are capable and increasingly reliable, particularly in constrained ODDs, but large-scale deployment should proceed alongside robust governance and public engagement. The technology’s promise is clear; achieving durable trust is a separate, equally challenging milestone.

Pros and Cons Analysis¶

Pros:
– High-fidelity perception from lidar improves detection accuracy and aids safe planning.
– Potential for significant reduction in road fatalities compared to human drivers over time.
– Transparent, data-driven operations can build trust through incident reporting and continuous improvement.

Cons:
– Public tolerance for machine-caused fatalities is extremely low, amplifying backlash risk.
– Edge cases and unstructured environments still challenge autonomous performance.
– Regulatory frameworks and liability standards remain fragmented and evolving.

Purchase Recommendation¶

For stakeholders evaluating investments in autonomous vehicle technology—OEMs, fleet operators, city planners, and mobility service providers—the guidance is to proceed with strategic prudence. Lidar-equipped platforms offer tangible safety and performance advantages, especially when integrated into robust sensor fusion stacks and conservative planning systems. These strengths make lidar a foundational component for any serious AV program.

However, deployment should be phased and data-driven. Prioritize geo-fenced operations where the environment is well-mapped and predictable. Establish clear performance metrics—incident rates, disengagements, near-miss analyses—and commit to transparent reporting. Build redundancy into sensing and compute, maintain rigorous maintenance protocols, and institute rapid incident response processes with accountable root-cause analysis.

Equally important is public engagement. Communicate capabilities and limitations honestly, avoid overpromising timelines, and collaborate with regulators to shape clear safety standards and liability frameworks. Prepare for crisis management before it is needed—because, as David Li cautions, even a single annual fatality linked to a company’s fleet can provoke existential scrutiny. Companies that anticipate this risk and design for resilience—technical, operational, and reputational—will be best positioned to deliver sustainable value.

In conclusion, invest in autonomous technologies that emphasize safety-first design, redundancy, and transparency. Lidar from established suppliers like Hesai is a strong foundation, but success hinges on the alignment of technology, governance, and societal trust. Adopt a measured rollout strategy, prove safety gains with evidence, and build credibility step by step. This disciplined approach maximizes long-term value while responsibly advancing the promise of safer, smarter mobility.

References¶

• Original Article – Source: techspot.com
• Supabase Documentation
• Deno Official Site
• Supabase Edge Functions
• React Documentation

*圖片來源：Unsplash*