[251028] #GM #Nvidia #RoboSense

Vol. 254 | 2025.10.28.

Category	Company	Article
LiDAR Sensor	GM	Cadillac Escalade IQ Getting Lidar-Powered, Eyes-Off Highway Cruising in 2028
Self-driving	Nvidia	How AI Is Unlocking Level 4 Autonomous Driving
LiDAR	RoboSense	The founder of the world’s top LiDAR maker says there’s one problem with Elon Musk’s approach to self-driving
Autonomous Trucking	Baidu	Baidu’s Apollo Go plans to launch taxis with no steering wheels in Switzerland as the race for robotaxis in Europe heats up
Autonomous Trucking	Kodiak	Kodiak Driver matches safest human fleets in Nauto safety rankings

Cadillac Escalade IQ Getting Lidar-Powered, Eyes-Off Highway Cruising in 2028

📌 GM will introduce a LiDAR-powered “eyes-off” highway driving system in the 2028 Cadillac Escalade IQ, marking a major leap toward Level 3 autonomy.

• GM announced a next-generation driver-assistance system enabling “eyes-off” highway driving, debuting in 2028 with the all-electric Cadillac Escalade IQ.

• The upgrade evolves from the current Super Cruise, which has logged over 700 million hands-free miles without a system-attributed crash since 2017.

• The new system adds a roof-mounted LiDAR sensor, combining LiDAR, radar, and camera inputs for enhanced redundancy and reliability.

• GM emphasized that multi-sensor fusion offers a major safety advantage over vision-only approaches used by some competitors.

• The Escalade IQ will feature turquoise interior and mirror lighting to indicate when full eyes-off mode is active.

• A new centralized computing platform will deliver 1,000× bandwidth, 35× higher AI performance, and 10× OTA update capacity compared to current systems.

• GM aims to start with predictable highway environments before expanding beyond, reinforcing its safety-first deployment philosophy.

• The 2028 launch marks GM’s next step toward scalable, LiDAR-enabled Level 3 autonomy across its vehicle lineup.

 

How AI Is Unlocking Level 4 Autonomous Driving

📌 AI breakthroughs such as foundation models, end-to-end architectures, and reasoning models are accelerating Level 4 autonomous driving, with NVIDIA leading the ecosystem from cloud to car.

• The SAE’s framework defines five levels of vehicle automation, with Level 4 enabling full autonomy within designated zones without human intervention.

• Over the past three to four years, progress toward Level 4 autonomy has advanced faster than in the previous decade.

• Six AI breakthroughs—foundation models, end-to-end architectures, reasoning models, simulation, compute power, and AI safety—are driving this acceleration.

• Foundation models provide world knowledge that allows vehicles to reason through unfamiliar scenarios, improving generalization and safety.

• End-to-end architectures connect perception, planning, and control within a single network, reducing complexity and improving decision-making.

• Reasoning Vision-Language-Action (VLA) models enable step-by-step, explainable decisions and enhance trust through real-time safety monitoring.

• Simulation and world-model technologies like NVIDIA Cosmos enable scalable testing in diverse virtual conditions, drastically shortening validation cycles.

• NVIDIA supports Level 4 autonomy with its full-stack ecosystem—DGX for training, Omniverse/Cosmos for simulation, and DRIVE AGX for real-time perception—forming a feedback loop for safer and faster deployment.

The founder of the world’s top LiDAR maker says there’s one problem with Elon Musk’s approach to self-driving

📌RoboSense founder Steven Qiu argued that a multi-sensor approach, including LiDAR, is essential for safe Level 3–4 autonomy—contradicting Elon Musk’s vision-only strategy.

• RoboSense founder and chief scientist Steven Qiu stated that a vision-only approach is insufficient for achieving safe Level 3 or Level 4 autonomous driving.

• He emphasized that multi-sensor fusion—combining LiDAR, radar, and cameras—handles “corner cases” that vision-only systems cannot manage.

• Qiu noted that a vision-only system might fail to distinguish objects such as a stopped white car or a black vehicle in a tunnel under varying lighting conditions.

• LiDAR costs have dropped sharply from around $70,000 per vehicle to just a few hundred dollars, while performance has significantly improved.

• Elon Musk has long criticized LiDAR as “expensive and unnecessary,” maintaining that vision-based autonomy will eventually make it obsolete.

• Other automakers disagree: Ford CEO Jim Farley called LiDAR “mission critical,” and Li Auto’s CEO Li Xiang said Musk’s view overlooks China’s more complex road conditions.

• The debate underscores a growing industry divide between vision-only and multi-sensor strategies in the race toward safe, scalable autonomy.

Baidu’s Apollo Go plans to launch taxis with no steering wheels in Switzerland as the race for robotaxis in Europe heats up

📌Baidu’s Apollo Go will begin robotaxi tests in Switzerland this December, aiming to launch fully driverless, steering wheel–free taxis by early 2027 as the competition for Europe’s autonomous mobility market intensifies.

• Baidu’s Apollo Go plans to start robotaxi testing in Switzerland in December 2025 through a strategic partnership with Swiss public transit operator PostBus.

• The companies aim to launch a public-facing driverless taxi service called “AmiGo” by the first quarter of 2027 using Baidu’s RT6 electric autonomous vehicles.

• Once the service is fully operational, the taxis will remove their steering wheels, signaling a move toward full Level 4–5 automation.

• The project marks Baidu’s first concrete step toward deploying robotaxis on European roads.

• Baidu is also expanding through partnerships with Lyft (U.K. and Germany in 2026) and Uber (outside the U.S. and China later this year).

• Competitors are accelerating their own European launches: Pony.ai will test in Luxembourg with Stellantis, and Waymo plans trials in London next year.

• Uber will also begin fully autonomous ride trials in the U.K. in 2026 with startup Wayve.

• The rapid succession of announcements underscores Europe’s emergence as the next major battleground for global robotaxi deployment.

Kodiak Driver matches safest human fleets in Nauto safety rankings

Click the image above to read the article.

📌Kodiak’s autonomous truck system achieved a top safety rating in Nauto’s AI-driven evaluation, matching the best human-operated fleets across the U.S. trucking industry.

• Kodiak AI’s autonomous trucking system achieved a VERA safety score of 98, tying for the top spot among more than 1,000 commercial fleets evaluated by Nauto.

• The VERA (Visually Enhanced Risk Assessment) score aggregates driving events and behavioral data into a single AI-based safety benchmark.

• Kodiak earned perfect scores (100) in inattentive driving, high-risk driving, and traffic violation categories, and 95 in aggressive driving.

• The result supports Kodiak’s claim that its self-driving trucks are among the safest drivers on U.S. highways.

• CEO Don Burnette said the company is combining internal safety analysis with independent, AI-powered validation to build public trust in driverless trucking.

• Nauto CEO Stefan Heck noted that Kodiak’s near-perfect score shows how AI can enhance fleet safety through real-time, data-driven insights.

• The evaluation strengthens the case for autonomous trucks as a safety-improving alternative to traditional human-driven fleets.

• Kodiak and Nauto plan to continue collaboration as Kodiak moves toward expanding fully driverless operations on U.S. public roads.

* Contents above are the opinion of ChatGPT, not an individual nor company

---

Tesla vs Waymo, Explained

---