Uber's Big Bet on Robotaxi: From Lucid and Nuro to Robo Express

Produced by Zhineng Technology

Uber, Lucid, and Nuro jointly invested over $300 million to advance the global deployment of 20,000 Level 4 Robotaxis, officially kicking off a new wave of commercialization competition for autonomous driving in the U.S.

With Tesla's Robotaxi initiative progressing, the U.S. Robotaxi industry is entering a critical phase of mass production, redefining the roles and collaboration logic among platform providers, vehicle manufacturers, and autonomous driving solution providers.

After comparing the technical approaches of Tesla and Waymo, Uber's choice of perception solutions appears more cautious and pragmatic. By partnering with China's Baidu Apollo, Pony.ai, and WeRide, Uber has effectively brought all major Chinese players into its fold.

Level 4 Robotaxi System Architecture:

Full-Stack Integration from Vehicle Platform to Perception Hardware

The Robotaxi project led by Uber involves Lucid's electric SUV Gravity platform, Nuro's Level 4 autonomous driving system, and Uber's fleet operation and dispatch network.

◎ The core vehicle platform is the Lucid Gravity, which builds on its existing high-performance electric architecture. Through a zonal control architecture and redundant chassis system, it reserves sufficient computing power, power supply, and signal path resources for Level 4 autonomous driving.

Lucid's vehicles feature an 800V high-voltage platform with high electronic and electrical bandwidth. The factory-equipped vehicles already include redundant steering, braking, and powertrain systems, laying the foundation for integrating advanced autonomous driving hardware and software.

Compared to traditional retrofitting, the "Robotaxi-ready" solution based on a pre-designed vehicle architecture is more efficient in thermal management, wiring, and controller deployment. It reduces system complexity and energy consumption while improving maintainability.

◎ For the autonomous driving system, Nuro provides its latest Level 4 solution in this collaboration.

According to disclosures, this solution is based on a lidar-dominant multi-sensor perception system paired with NVIDIA Thor central computing platform. The system deploys at least four lidars, multiple 8M cameras, millimeter-wave radars, and ultrasonic sensors to achieve 360-degree blind-spot-free perception in all weather and scenarios.

The Thor platform delivers over 1000 TOPS of AI computing power and supports parallel processing of multiple tasks, including semantic segmentation, path prediction, sensor fusion, and decision control. Nuro's software stack is a neutralized Level 4 platform refined from its unmanned delivery business, already commercially validated in low-speed closed environments and now expanding to open-road scenarios.

To ensure system safety, the vehicle adopts a triple-redundancy mechanism—automatically switching to backup channels if the main control system fails. The braking and steering control units are dual-control designed, with an independent communication bus to prevent CAN network conflicts.

While such designs increase costs, redundancy and stability are prioritized over functional complexity for Robotaxi operations, which rely entirely on system decisions without human intervention.

From a technical perspective, Uber's Level 4 Robotaxi, developed in collaboration with Lucid and Nuro, is no longer a mere stacking of sensors and algorithms but a full-stack integration spanning underlying vehicle architecture, system redundancy, perception computing power, and platform dispatch.

◎ Lucid's high-voltage platform and zonal controller design provide ample space for hardware expansion;

◎ Nuro's multi-sensor fusion and redundant control systems emphasize safety first;

◎ Uber is responsible for embedding this complex system into a city-level operational network, laying the groundwork for global deployment.

Divergence and Validation of Technical Paths:

Why Does Uber Favor a Fusion Approach?

Uber's strategic focus is not on building its own system but on collaborating with proven solution providers to accelerate development.

Amid the growing hype around the Robotaxi market, this "platform + external technology integration" approach speeds up deployment and reduces early-stage R&D risks. Uber's simultaneous partnerships with Nuro and Baidu Apollo also highlight its preference in key technical paths.

Current Robotaxi perception technologies broadly fall into two camps:

◎ One is Tesla's vision-only approach, emphasizing cost control and high integration;

◎ The other is the multi-modal fusion approach (lidar + cameras + millimeter-wave radar) championed by Waymo and Baidu Apollo, prioritizing system robustness and safety redundancy.

Data from California's DMV shows Waymo's fusion approach has significantly lower disengagement rates than Tesla's vision-only system. Lidar's reliability and accuracy are particularly superior in nighttime, adverse weather, and complex road conditions.

For example, lidar achieves 98.5% accuracy in nighttime pedestrian detection, compared to 82.3% for vision-only systems.

Baidu Apollo's RT6 is a prime example, featuring four 128-line lidars with a 200-meter range and over 1.53 million points per second, alongside twelve 8MP cameras, five millimeter-wave radars, and twelve ultrasonic sensors, forming a five-layer 360-degree perception system.

This multi-redundancy architecture, though costlier, significantly enhances Level 4 autonomous driving's ability to handle complex urban roads and unexpected scenarios (e.g., construction zones, temporary detours).

Baidu Apollo also incorporates large models, with its Apollo ADFM model capable of intent recognition, behavior prediction, and path coordination in dynamic traffic environments, further improving Robotaxi adaptability. This fusion of traditional perception stacks and AI models is becoming an industry trend.

Uber's choice to partner with Nuro and Baidu Apollo stems from the current limitations of vision-only systems in extreme conditions. Lidar-based multi-modal perception, despite higher costs, offers clearer safety margins and stronger fault tolerance, making it ideal for large-scale Robotaxi deployment. These solutions also have proven experience in multi-city, multi-climate operations, providing replicable models for Uber's global network.

Summary

Uber's moves consolidate global players into a comprehensive game of platform, hardware, and software synergy centered on vehicle intelligence.

From underlying electronic architecture to perception stacks and operational dispatch, Uber has built a Robotaxi deployment system prioritizing stability and safety. China's engineering capabilities, system integration, and operational maturity in high-level autonomous driving are gaining recognition from mainstream platforms.