Autonomous Robot Delivers Takeout Using Multi-Component Vision System

What would you think if a robot using a sophisticated vision system could deliver a pizza?

That may soon be a reality thanks to research by Starship Technologies. The London-based company envisions a future where autonomous robots can make many different types of deliveries. Their concept was recently tested with outstanding results.

A UK-based food delivery service, Just Eat, used an autonomous robot to make a successful delivery from a Turkish restaurant to the consumer. The order of falafel and lamb cutlets crossed town safely from the restaurant to the hungry customer in a locked compartment.

The Multi-Component Machine Vision System is The Key to Success

Many different systems worked together to achieve this groundbreaking result, but it would not have been possible without a particularly sophisticated machine vision implementation.

The intrepid delivery robot incorporates nine visible cameras with three pairs configured for stereo vision. It also uses multiple “time of flight” (TOF) cameras. An Nvidia Tegra K1 mobile processor facilitates integration of vast quantities of machine vision input into the robot’s autonomous driving.

Naturally, the robot is outfitted with powerful GPS. The on-board GPS is accurate to about 30 meters, but this was not considered precise enough for the fast-paced operational environment. Thus, the system is augmented with algorithms that build a 3D map of the neighborhood. By combining input from machine vision and ultrasonic sensors, the robot navigates this map in a “personal awareness bubble.”

A Bright Future for Autonomous Robots All Throughout Logistics?

At 40 pounds and with a carrying capacity of about 22 pounds, the robot has vast potential to make all kinds of deliveries – not just food items. Ultimately, similar robots are intended to become "personal couriers" that can make deliveries from all local shops.

So far, they are off to an auspicious start!