“Our delivery robots need a platform that has significant computing power and can interface with different types of sensors to safely drive outdoors in dynamic and unstructured environments”
Chinese e-commerce companies JD.com and Maituan will use NVIDIA’s Jetson AGX Xavier platform to underpin their next generation of autonomous delivery robots.
Currently China’s largest retailer JD.com uses a delivery robot to fulfill customer orders in selected cities. The driverless vehicle moves at 15km per hour and has a claimed carry capacity of 300kg; far above that of rival robots being trialled elsewhere.
As the technology is still in a nascent stage the robots only deliver to residents within a “reasonable distance” of JD’s distribution centres.
Autonomous delivery vehicles use LIDAR sensors and high-resolutions cameras to navigate around their environment. Bustling city surroundings contain numerous hazards and obstacles for the robot delivery vehicles ranging from cars, objects, pedestrians and increasingly other robots. In Beijing special non-motor-vehicle lanes are in place so that the driver-less vehicles can run unobstructed.
Qi Kong, who leads development of JD.com’s robotics research said in a NVIDIA blog that: “Our delivery robots need a platform that has significant computing power and can interface with different types of sensors to safely drive outdoors in dynamic and unstructured environments, along with interacting with humans.”
JD’s current generation of robots can carry up to 300kg, but this is something the company expects to be scaled down slightly as they finish the testing period and strive to make the robots safer.
Jetson AGX Xavier
Jetson AGX Xavier was designed by NVIDIA for the robotics market.
Modern Robotics are equipped with an array of sensors: in order to run these the Xavier contains six high-performance processing units.
Its GPU is a 512-core NVIDIA Volta architecture with Tensor Cores. While its CPU is an 8-core ARM v8.2 64-bit. The system is capable of delivering 32 deep learning TOPS or a trillion operations per second.
It holds 32GB worth of memory on an embedded multimedia card.
Running the sensors while also undertaking the calculations required to avoid hazards are not light tasks when it comes to energy consumption. NVIDIA claim that the Jetson AGX Xavier is ten times more energy efficient than its predecessor, the Jetson TX2. The unit can operate at 10W, 15W or 30W as needed.
Autonomous delivery technology is been tested and rolled out by several companies over the last few years and consumers living in large population centres should expect to encounter the technology. London-based start-up Starship Technologies aimed to deploy 1,000 delivery vehicles on a California campus this year.
The robots were tasked with delivering office supplies between work stations and acting as a delivery services for local food stalls.
The delivery vehicles run at speeds of 4mph and are currently in commercial use in Bern, Hamburg, Milton Keynes, San Jose, Talinn and Washington DC.