Nvidia (NASDAQ: [[ticker:NVDA]]), the Santa Clara, CA-based maker of specialized graphics processing chips that have become workhorses in artificial intelligence, has joined a group of investors that put $20 million into TuSimple, a Chinese startup developing self-driving technology for long-haul trucks.
In a blog posted Wednesday, Nvidia says TuSimple has used Nvidia GPUs, the Nvidia Drive PX 2 computing module, and other devices and technologies to develop its autonomous navigation system.
The race is on among automakers, technology companies, and startups to assemble a suite of sensors, software, and infrastructure that will allow vehicles to drive themselves safely without human intervention. Earlier this year, Nvidia rival Intel announced a $15.3 billion acquisition bid for Mobileye, which makes cameras and software for autonomous vehicles. Last week, Xconomy reported on a new PARC-spinout, Metawave, which is building metamaterials-based radars backed with artificial intelligence software.
Nvidia, founded in 1999 and based in Santa Clara, CA, has expanded beyond its initial focus on graphics processing units (GPUs) used in gaming and 3D visualization to become a leader in AI technology required in everything from data centers to automobiles. The company’s GPU-based approach to machine learning is targeting problems in cancer detection, weather prediction, and self-driving vehicles. Nvidia’s competitors include Intel (NASDAQ: [[ticker:INTC]]) and San Diego-based Qualcomm (NASDAQ: [[ticker:QCOM]]).
Founded in Beijing in 2015, TuSimple has about 100 employees in China and the United States.
In an e-mail Thursday, TuSimple spokeswoman Zhujia “Claire” Shi said the startup operates a second R&D center in San Diego with over 30 employees. “We chose SD because first, the operation cost is relatively low, secondly, there is a good autonomous driving infrastructure and government support from SANDAG and many more.”
TuSimple worked through the San Diego Association of Governments (SANDAG), a regional transportation planning agency, to prepare for a 200-mile test drive of a tractor-trailer rig equipped with its self-driving technology. The road trip from San Diego to Yuman, AZ, relied on Nvidia’s technology, camera arrays, and millimeter-wave radar to scan more than 656 feet down the road—equivalent to the length of two football fields.
TuSimple said its test drive in June met SAE Level 4 standards for the automation of on-road vehicles. SAE, an international society of professional engineers once known as the Society of Automotive Engineers, defines Level 4 as “High Automation: The driving mode-specific performance by an Automated Driving System of all aspects of the dynamic driving task, even if a human driver does not respond appropriately to a request to intervene.”
TuSimple says its technology includes extensive high-definition mapping of the routes driven, as well as computer vision processing of real-time data from three millimeter-wave radar units and eight on-board cameras. TuSimple says it can achieve “centimeter-level” accuracy for truck positioning and its in-house decision-making machine intelligence makes for safe route navigation.
The startup says its deep learning algorithms enable its onboard auto-navigation system “to perceive the surroundings just like the human eye. It can detect and track objects within your field of vision in real time, and make pixel-level interpretations of the visible scene.”
TuSimple also has been developing technology that uses image recognition software to identify the makes and models of cars around its self-driving trucks.
