Bibhrajit Halder honed his AV street cred with Caterpillar, Apple, Ford and the DARPA Grand Challenge before he pivoted to autonomy for mining and construction vehicles. It’s been a good move, he related in an interview with Fierce Electronics.
As founder and CEO of SafeAI, based in Santa Clara, Calif., his team retrofits massive heavy equipment with AV hardware and software. The six-year-old company has publicly named a large construction company as well as a mining customer using the company’s products, and is working with other companies in the sector he will not name.
Japanese construction company Obayashi started working with SafeAI in 2020 and last year signed an expansion of their deal that brings in Siemens to create an autonomous electric fleet.
Also last year, SafeAI signed an expansion of a deal with Australian mining company MACA for a 100 truck fleet.
Companies can use SafeAI products to retrofit an existing fleet rather than starting from scratch. The platform is vehicle agnostic and relies on multimode sensors and powerful onboard processing that runs on the SafetAI Autonomous Framework OS, which was purpose built for autonomous heavy equipment.
On-board sensors include a mix of cameras, radar and lidar analyzed to dynamically map the terrain and detect and react to obstacles in real-time. Depending on the size of the vehicle, SafeAI installs three lidars with two in front and one in back, two to four cameras and two to four radars. V2X is embedded for personnel and vehicle communications and safety, according to the company’s web site, helping trucks be aware of other vehicles whether manned or autonomous.
At first, worksites are mapped using lidar and GNSS, while perceptions allow vehicles to recognize where they are relative to the underlying map and also adapt as environments change. The system’s AI determines how the vehicle executes required activities and the onboard computer uses SafeAI software to a drive-by-wire system to control driving and operations.
While mining and construction pose tremendous safety challenges with steep cliffs and narrow passages requiring precise control, Halder said the challenges have actually been less demanding that creating AVs for use on public roadways. “Mining and construction use cases involve repetitive work, 24x7, and autonomy is a no-brainer because it is such a constrained environment without passengers,” he said. That's not to say its exactly easy to build autonomy for these big rigs, just that it's relatively easy when compared to self-driving cars on public roadways. On these work sites, “everything is very well organized and a normal human won’t walk into a site which is well-controlled with a geo fence.”
With SafeAI there is no remote control of trucks and no human operator. The company sells a software license and with an annual fee. All in, he said a construction or mining company can save 25% on their trucking costs. So far, it seems the sky is the limit in terms of potential interest.
“Our biggest challenge is executing as fast as possible,” which primarily means finding the right engineers for the work. So far, the company has 98 employees.
There have been no accidents over the past eight years across hundreds of AV trucks used in the industry, he said. “Safety is number one,” he said.
Several well-known large mining truck manufacturers and a few small startups provide various levels of autonomy for their vehicles. Halder said his company’s competitive advantage is knowing how the big companies operate, partly due to his seven years at Caterpillar. One smaller UK-based company, Oxbotica, is a potential competitor and markets itself as offering universal autonomy for any vehicle.
RELATED: Gigantic and smart machines took the CES 2023 stage