Ex-Apple car engineers’ startup Vayu offers autonomous delivery robots sans Lidar sensors

Generative AI is ushering in a new age of smart robots that can handle complex workflows in enterprise settings. Several players are working in this domain, but it is just the beginning. Soon, this technological leap may even reach your doorstep, thanks to California-based Vayu Robotics and its novel on-road delivery robots.

While multiple organizations, including Kiwibot, Neolix, Serve Robotics and Starship Technologies, have explored the idea of automating deliveries with robots, Vayu stands out with its unique approach. The company has tapped the power of foundation models to create a delivery system that can navigate autonomously in any environment — without using the complex and expensive tech stack normally required.

“The unique set of technologies we have developed at Vayu has allowed us to solve problems that have plagued delivery robots over the past decade, and finally create a solution that can actually be deployed at scale and enable the cheap transport of goods everywhere,” Anand Gopalan, the founder and CEO of the company, said in a statement. 

The company has been around only for a little over two years but is already getting traction from enterprises, including large ecommerce players looking to deliver on the promise of ultra-fast product deliveries.

Eliminating expensive Lidar-based tech stack

Mobility systems have long relied on a Lidar-based tech stack. The sensors emit laser pulses and measure the time it takes for the light to bounce back from objects, producing information for different algorithms to map the environment in 3D and enable the robot to navigate it effectively.

The approach works well but often hits enterprises with one major issue: high costs. Essentially, high-quality lidar sensors, especially those with long-range and high resolution, can be costly and supply constrained. This can significantly increase the overall cost of building robotic systems, making them difficult to scale. 

Gopalan, previously the CEO of Lidar sensor company Velodyne, saw this gap when dealing with potential customers in the robotics and autonomous mobility industries.

According to him, the customers built their first systems successfully but ended up hitting a wall because either the hardware was too expensive to scale or the accompanying software couldn’t deal with uncertainties (things like fog or direct sunlight hitting the sensor). 

By contrast, Tesla Motors has pursued driver assisted “self-driving” without Lidar as well, and without radar (sound-based detection), using a computer vision only approach that was initially viewed skeptically by rivals and experts, and even some within the company, but has now proved to be among the more accessible, cost-effective and resilient self-driving modalities to date.

Now it appears that Gopalan and his colleagues at Vayu have also embraced a similar approach.

Rising from the ashes of the scuttled Apple Car

Prompted by the need for robots that are scalable and cost-effective from the get-go, he started Vayu Robotics with Mahesh Krishnamurthi and Nitish Srivastava from Apple’s now-scrubbed Special Projects Group, focused on the self-driving “Apple Car,” reportedly canceled in February 2024.

The trio started operations in 2021 and eventually developed a ready-to-scale autonomous delivery robot that can navigate complex environments, including city streets and shop floors, at a speed of less than 20mph while carrying payloads weighing up to 100 pounds. 

The biggest differentiator: the Vayu robot does this without using expensive Lidar sensors or pre-mapping the environment it has to navigate. Instead, it uses a transformer-based “mobility foundation model” with a new type of passive sensor.

“The mobility foundation model has been trained in a simulation environment developed by Vayu. The environment allowed us to procedurally generate millions of moving miles and challenging corner cases. Vayu has also developed a unique sim-to-real transfer approach that allows us to transfer the learning from this AI model to the real world. This enables us to create an autonomy stack with a fraction of the time and compute cost of a traditional approach,” Gopalan explained.

Flexible tech

The CEO noted that the mobility technology is neither form factor nor application-exclusive. It can be used in different fields of work and even replace existing Lidar-based tech stacks across robotic systems.

The company chose to start with delivery robots as ecommerce is witnessing a massive surge but the cost per delivery continues to remain very high. Vayu’s autonomous systems, having the same cost basis as a standard automotive camera module, can easily bring that down. 

“The delivery robot’s system will connect to the customer’s ecommerce order dispatch workflow, and receive the destination location and order specifics similar to a third-party logistics provider. Our first customers usually house the robots at their warehouse or dark store location and are responsible for loading the physical goods into the robot. Once that’s done, Vayu’s autonomy system will take over. It will travel to the specified end location, drop off the package and proceed,” he added.

Over 100 prospects on waitlist for Vayu robot

While Vayu started commercialization efforts for the delivery bot only a few months ago, it is already witnessing significant traction, with as many as 20 enterprises piloting the technology and over 100 prospects on a waitlist. The company also recently signed a commercial agreement to deploy 2500 robots for a large ecommerce company.

“In our test phase, the robots have clocked up thousands of miles on the road and millions virtually. With a high degree of confidence, we anticipate growth to $50M over the next 3 years. If we include a broader look at the market opportunity for both robotic delivery and low-cost depth sensing, we could easily exceed this amount,” Gopalan said.

The company also plans to demonstrate how the technology can go beyond wheeled robots and drive the function of bipedal and quadrupedal robots. It will showcase the solution by the end of this year.