From an article in New Atlas by Paul Ridden.
We’ve seen a bunch of delivery robots trundle around campuses and neighborhoods over the years, with companies like Amazon, Fed-Ex, Walmart, Uber Eats and others all trying out different solutions. A Californian startup is rolling down a somewhat different path, with an autonomous bot that uses a new low-cost vision system plus AI training to navigate without pre-mapping the route, and does away with pricey sensor suites to boot.
Vayu Robotics was founded in 2021 by “engineers, technologists and business leaders with decades of industry experience developing and commercializing cutting-edge automotive sensing, autonomous vehicles and robotics technology.”
The team launched a newly developed camera sensor in 2022, aimed at allowing its autonomous delivery bots to roll without LiDAR sensors. Vayu Sense “combines dense, low-cost CMOS image sensors with modern computational imaging and machine learning techniques.” The company claims that this proprietary technology not only outperforms typical RGB cameras but LiDAR too, resulting in a cost-effective high-resolution robotic vision system with high-res depth perception, object detection and the ability to work effectively in challenging conditions.
This was followed by a proprietary foundation AI model for robotics autonomy called the Vayu Drive that’s trained using both simulated and real-world data, and negates the need for HD maps, localization technology or LiDAR – relying on the Sense vision system instead.
“It is an end-to-end neural network, similar to LLMs in that it operates on a tokens in and tokens out basis,” explained the company. “The input is multimodal – image tokens from the cameras, instruction tokens for the instruction that the robot has been told to perform, route tokens to show it the road-level navigation path.
For the complete article on the new delivery robot CLICK HERE.
