Waymo Engineering Exec says that the Lidar and Radar sensors are the key to safety in the autonomous technological battery

“We have done a lot of research. We are aware of what works and what does not work on our scale and what we have to do,” said Srikanth Thirumalai, vice-president of engineering integrated into the Robotaxi industry, Waymo, said this week at the AI4 conference in Las Vegas.

While rivals like Tesla push autonomous cars that depend only on video cameras, Thirumalai de Waymo says that the combination of Lidar and Radar provides “an additional safety net” to ensure that the company has the appropriate data which it needs to make driving decisions “in all conditions” – including extreme times.

Thirumalai spoke on stage in an interview with Fortune. Earlier in the day, Thirumalai made a solo presentation, describing in detail the AI of Waymo battery and the approach of security that allowed the company to bring its operation to five cities by mid-2025 and perform more than 100 million kilometers without driver. In his presentation, Thirumalai showed a video of the way Lidar sensors on Waymo Jaguar I -Pace had taken the movement of human beings preparing to jump on the road, even when the cameras of the vehicle had not done it – or a woman preparing to go around a bus stopped and directly in the way of a Waymo Robotaxi. In both cases, the Waymo Robotaxi stopped or maneuver to avoid contacts with pedestrians, depending on the videos.

The presentation showed the striking contrast in the approaches between Waymo and one of its new rivals, Tesla, which launched a small -scale Robotaxi service only in Austin in June, with safety drivers in the passenger seat. Tesla, which demonstrated its full technology on self-driving (FSD) via demonstration walks at the AI4 conference, only uses video cameras and its AI technology for FSD and Tesla Robotaxi, after years of Elon Musk, declaring that other sensors are expensive and invented. “The Lidar is a crazy race,” said Elon Musk in 2019. “Anyone composed on Lidar is condemned. Convicted! (These are) expensive sensors that are not necessary.”

Thirumalai would not say directly if he considered the autonomous camera systems as Tesla to be sure for public roads. He said that you must consider “the whole process” of how a system is built, tested, then validated, and he also said that you could not statistically compare the Waymo system to another, due to the lack of comparable security measures. General Motors’s subsidiary, which also used Lidar and Radar systems, suspended operations earlier this year after failing to relaunch after a serious accident in San Francisco. For the context, Tesla said he had traveled 7,000 miles without driver at the end of July, against 100 million Waymo.

“If we are talking about objective measures, then we must examine the statistics of our security file, on a large scale, right?” Said Thirumalai. “When someone really says: Yes, we have equaled your safety on your scale with a different system, it’s great. We will take this. ”

Waymo regularly tests new technologies as it becomes available, according to Thirumalai. As part of this experiment, he declared that Waymo sought how multimodal models such as Gemini can be incorporated into the Waymo technological battery (Waymo has not tested any other generative AI model in addition to Google Gemini, confirmed Thirumalai). The company Robotaxi has published several articles of its research on multimodal models, including a traffic simulation on the city scale with a generative world model as well as the search for Waymo on Emma, the multimodal model from end to end of Waymo for autonomous driving. Waymo reported that the co-training of its vehicles with Emma has helped things like the detection of objects and road graphics, claiming that there was “potential” for Emma as a general model for autonomous driving applications. However, Emma is expensive, can only deal with a small number of image frames and does not incorporate Lidar or radar sensors, which leads to “challenges” for the use of Emma as “autonomous model to drive”

Thirumalai said that the integration of generative AI models in the autonomous technological pile is an “intense research” area and that it will continue. “But there is much more work that will be necessary to make the system as simple as possible,” he said.