The framework for autonomous delivery robots in the coming years is already in place. Although some major delivery companies have scaled down development, many specialist robotics companies are on the verge of scaling their successful proof-of-concept (PoC) deployments in different geographies. But with more and more delivery robots on our pavements, cycle paths, and roads, how can we ensure they don’t go rogue - behave erratically or dangerously - especially by disregarding the local traffic rules?
Here we look at the unassuming robotic lawnmower (RLM), already widely used worldwide, and consider how these platforms can be adopted for autonomous delivery robots. We also look at the challenges ahead that must be overcome to ensure that these delivery robots operate safely and reliably and, more importantly, in harmony with pedestrians and other road users.
Global Navigation Satellite Systems (GNSS) and other localisation technologies enable the precise location of objects and people almost in real time.
However, for sectors such as mobile delivery robotics, locating obstacles is not enough; they need to know their precise position too. Mapping the surrounding space by creating a virtual geofence enables them to determine the spatial limits of a specific area and the physical obstructions nearby.
Geofencing relies on two key technologies: Real-Time Kinematic (RTK) positioning and dead reckoning (DR). RTK corrects common errors in current satellite navigation (GNSS) systems, providing better positional accuracy down to the centimetre level. DR calculates the current position of a moving object by determining an already fixed position within the virtual geofence and then incorporating estimates of speed and heading direction over time.
The robotic lawnmower is one of the most tangible examples to illustrate how GNSS/location technologies have developed over the years.
The beginnings of geofencing
Robotic lawnmowers are leading developments in the field of geofencing. Back in 2011, these lawnmowers relied heavily on mapping technologies using hall-effect, lidar, radar and other proximity sensors.
Since then, however, technological developments have focused on establishing a geofence to determine the shape of an area, such as public parkland or golf course, and enabling navigation within an established perimeter while excluding trees and static obstacles.
In the early days, the perimeter of the area was physically marked out using a wire. This technique was effective but not ideal as any breaks in the boundary wire rendered the system useless. There were also the installation cost and the cost of the wire to consider. A few years later, RTK, using a local base station to provide the GNSS error corrections, enabled virtual perimeters.
The next step will involve removing the local base station and relying only on error corrections provided by a network of base stations already deployed in the field. This will bring further ease of use and platform portability benefits.
Delivery robots are a similar use case to RLM for geofencing technologies but slightly more complicated because they need to determine their surrounding space in seconds. Detecting moving objects to avoid crashes adds further complexity. Lidar and radar technologies currently achieve this, but that’s not the end of the story.
Challenges ahead
Challenges, not evident at first sight, are now revealing themselves that hamper the wide-scale deployment of autonomous delivery robots.
The first challenge to overcome is the ad-hoc policies on local, regional, and national road traffic rules for delivery robots. The time when governments only issued road traffic laws for one or possibly two types of vehicles is long gone. In addition, delivery robots can travel along sidewalks and cycle routes at various speeds and in various directions.
As a result, traffic is out of control, and pedestrian and other motorist safety are all at risk. Issues include: where are these delivery robots allowed to circulate, and what is their maximum speed according to location?
RTK-based high-precision positioning, such as the system used for RLMs, answers many of these new challenges. These systems can enable cm-level geofencing required in use cases such as no-go zones in pedestrian areas, banned pavements and bicycle lanes, and speed limits supported by geofencing. However, this reveals a second challenge for delivery robot manufacturers. Ideally, they want to develop a platform that works seamlessly around the globe, but there is a lack of standardisation across the various available GNSS error correction services.
In order to create an open communication standard, a consortium of GNSS service providers and equipment manufacturers, including u-blox, established the Safe Position Augmentation for Real-time Navigation (SPARTN) format. Introduced in January 2020, this industry-recognised standard supports wide and global area broadcasts.
Japan’s delivery robot rollout
The Level 4 public road driving ban was lifted in Japan on April 1, 2023. As a rule of thumb, the more historical the cities, the narrower the roads. Municipal roads make up 80% of all roads in Japan, according to the Ministry of Land, Infrastructure, Transport, and Tourism, and they are typically only 3.8 meters wide. The risk of accidents, such as contact with various obstacles and other vehicles, is higher than in major Western countries.
In the run-up to the enforcement of the revised road traffic law in Japan, the Ministry of Economy, Trade and Industry held an event to illustrate the work it has been doing in collaboration with the Robot Delivery Association. The ministry believes that it is important to let other passers-by know about the delivery robots when they run on sidewalks, which are public spaces, and plans to work on improving social acceptability.
Now, robots that meet certain size and structure requirements, equivalent to the size of an electric wheelchair, will be able to run on public roads and sidewalks at a maximum speed of 6kmh (just under 4mph).
However, there is an obligation for the delivery robot operator to notify the prefectural public safety commission in advance. At the event, eight delivery robots from eight companies demonstrated actual driving and usage scenarios. These included basic movements, such as going straight and turning, a person jumping in front of the robot and the robot suddenly stopping, unlocking a locker, and taking out products.
Conclusion
This combination of ultra-high precision GNSS solutions, an extensive network of base station antennas, online data services, and standardised error correction services, such as SPARTN, will help accelerate the rollout of delivery robots across the world. Used smartly, all the building blocks should be in place to bring operators and users alike benefits such as increased efficiency and lower service costs.
Today lidar and radar technologies are widely used for determining the delivery robots surrounding space and detecting moving objects in almost real-time. In the coming years, Vision-RTK promises even better results. The latest generation Vision-RTK2 fuses computer vision, RTK-GNSS, inertial sensors, and odometer data from its wheels with machine learning algorithms - no longer depending on GNSS signals alone.
This technology is especially interesting in challenging environments like urban canyons, underpasses, tree canopies, or anywhere else where satellite signals can be blocked, shielded, or intervened by reflection.
Author details: Diego Grassi, Head of Segment Managers Industrial, EMEA, u-blox
