VayuRobotics today released its first delivery robot. One can follow a clerk as they load customers' orders in a store, then autonomously navigate city streets at speeds of up to 20 mph to deliver the goods.
Over the years, companies like Amazon, FedEx, Walmart, UberEats, and others have tried different solutions, and we've seen many delivery robots crisscrossing campuses and neighborhoods. A California startup is going down a slightly different path, with its autonomous robot using a new low-cost vision system and artificial intelligence training to navigate without pre-drawn roadmaps and eliminating expensive sensor suites.
VayuRobotics was founded in 2021 by "engineers, technology experts and business leaders with decades of experience developing and commercializing automotive sensing, autonomous vehicles and robotics technologies."
The team unveiled a newly developed camera sensor in 2022 designed to allow its autonomous delivery robots to roll without lidar sensors. VayuSense "combines high-density, low-cost CMOS image sensors with modern computational imaging and machine learning technologies." The company claims that this proprietary technology outperforms not only typical RGB cameras but also LiDAR, resulting in a cost-effective, high-resolution robotic vision system with high-definition depth perception, object detection, and the ability to work effectively in challenging conditions.
Subsequently, the company also launched "VayuDrive", a proprietary basic artificial intelligence model for robot autonomous driving, which is trained with simulated data and real-world data without the need for high-definition maps, positioning technology or lidar, but relies on the Sense vision system.
The company explains: "It is an end-to-end neural network, similar to an LLM, that operates as labeled inputs and labeled outputs. The inputs are multimodal - image tokens from the camera, command tokens that instruct the robot to perform instructions, route tokens that show the robot a road navigation path."
Unlike other LLMs, it has a "state" concept that is built over time and updated with each frame of additional input received. This allows the use of large context windows without the typical slowdown seen with large contexts. It's designed to run efficiently at the edge at 10 frames per second.
VayuRobotics, which emerged from stealth last October with $12.7 million in seed funding from backers including Lockheed Martin, has announced the launch of its first delivery robot. One can walk on roads, bike paths, sidewalks and in shops, and is the first of its kind to use artificial intelligence-based models and low-cost passive sensors.
The four-wheeled electric delivery pod is 3.3 feet (1 meter) high, 5.9 feet (1.8 meters) long and 2.2 feet (0.67 meters) wide, so it won't be too much of an obstacle to other vehicles when making deliveries at speeds of up to 20 mph (32 km/h). Its battery pack reportedly has a maximum range of 60 to 70 miles (up to 112.6 kilometers) per charge.
After arriving at the drop point, it can park on the sidewalk or driveway, open the side door, and use its robotic arm to take out the designated package. The company says its storage compartment can hold up to 100 pounds (45 kilograms) of cargo, but with some adjustments that can be increased to 200 pounds.
One is currently being tested by an as-yet-unnamed "large e-commerce company," which plans to deploy 2,500 robots, starting in San Ramon, California, and then expanding to other cities across the United States. Other commercial customers are expected to join the program, but Vayu also hopes to use its technology in other robotics applications and is currently working with a "leading global robot manufacturer" to replace the lidar sensors with Vayu sensing technology.
"Our software is robot form factor agnostic and we have deployed it on multiple wheeled form factors. In the near future, Vayu's software technology will enable quadruped and biped robot mobility, allowing us to expand into these markets as well," revealed company co-founder Anand Gopalan. Watch the video below for more.