Hermes delivers robot couriers onto the streets of London

Courier company Hermes is trialling autonomous, self-driving electric robots to deliver parcels to customers.

The company has partnered with Starship Technologies, the robotics start-up co-founded by Skype pioneers Ahti Heinla and Janus Friis, to test half a dozen of the robots in London. Hermes has carried out 10,000 miles of testing to date in the capital’s streets, including during Easter week in busy Bermondsey Street in Southwark. The programme follows similar tests by Starship in Hamburg last summer.

Hermes says that the project is to designed to explore new on-demand delivery models, inspired by the lives of time-poor millennial customers who want instant gratification, according to Hermes’ Head of Innovation David Turner.

So why invest in street-level robots rather than Amazon- or DHL-style airborne drones? Turner says:

Drones seem to have a good fit with remote areas, where it takes time to get deliveries out there – for example, medical supplies to remote islanders. But we felt that there were more restrictions on drones in built-up areas.

But there remain serious obstacles to any autonomous vehicle, admits Turner, including existing legislation, not to mention local councils’ differing attitudes to driverless machines in their communities. This is one reason why the Starship robots are accompanied by human walkers (until legislation changes), while remote intervention teams can also take control of the robot if its sensors detect problems.

This begs the question of why Hermes can’t simply give the parcel to a human walker to deliver – but this is a testbed technology, designed to explore the new business models that will emerge once the law has caught up with the innovations. And once self-driving technologies have proven themselves to be safe: any vendor or client could be a single collision, death, or injury away from a devastating lawsuit. Where will responsibility lie?

So why else might Hermes employ self-driving, autonomous machines? While most companies automate primarily to save money, Turner explains that this is not the case with Hermes, which is exploring new delivery models to meet emerging needs – with the focus being on the 15-minute time slots rather than on the robots themselves. He says:

We’re not looking at the cost aspect, we’re looking at the customer experience. First of all, this is about the ability to extend our portfolio to offer an on-demand, 15-minute service to our customers and see if they like it.

Yet, surely, the money-saving attractions of autonomous/self-driving vehicles may prove irresistible to any delivery company in the future. Turner admits that Hermes is talking to a range of providers about their offerings as part of its experiments with new service types, in a world in which Tesla, Jaguar/Land Rover, Mercedes, and Volvo (not to mention Google and Apple) are all accelerating in the same direction:

As regards the autonomous part, I think we’re talking to a couple of companies about self-driving vehicles, but a lot of this is purely experimental in terms of what this future will look like, and how it will work.

Another impetus behind the experiment is the changing attitudes to vehicles in urban areas, says Turner, where emissions targets and population growth will make the increased demand for electric vehicles inevitable:

This is also about looking at the strategic future for inner-city deliveries, where we have to reconsider our vehicle strategies. Councils will clamp down on some kinds of vehicles, so it would be remiss of us not to look at electric and autonomous deliveries. Plus, automation allows us to be more intelligent with our deliveries. It would be problematic for our human couriers to fulfil these 15-minute slots.

Augmentation

That said, this is a case of augmenting Hermes’ service, and not of replacing human couriers with robots, he claims. In the future, the same location-based technology platform that Hermes is looking to build could be used to offer rapid delivery slots to Hermes’ human workforce, too – if they are close enough to a target address and wish to accept the job. Turner explains:

What this on-demand service allows us to do, by moving data in real time, is to deviate parcels into a parcel shop, so if the human courier is nearby, via GPS, we could ask that courier to deliver in the next 30 minutes if they are, say, 0.6 miles away from the shop.

We’re trying to get clever with our data. We’re looking at optimisation based on where the assets are at any one time. So rather than playing on the ‘courier or robot?’ side of things, we’re simply experimenting with our product.

At present, Hermes is leasing the robots rather than building a fleet of them – in the same way as train operators lease rolling stock. Other companies, such as Just Eat and Domino Pizza are also using Starship Robotics’ machines. Meanwhile, the small number of them in the Hermes trial means that the solution doesn’t scale in its current form, according to Turner:

We’re talking about 5 or 6 robots at the moment, a few parcels a day. A busy Oxford Street filled with our robots just isn’t going to happen.

But in the long term, it stands to reason that while the increased use of drones and self-driving vehicles may open up new options for data-rich, time-poor millennial customers, it also shifts some of the onus and responsibility onto that customer, and away from the courier.

For example, if they live in a block of flats, most people would have to go down to the street – perhaps some distance from their home – to collect their parcel from a robot, rather than have it delivered to their door. This would be of little benefit to elderly or disabled people, even when time is critical. So might face-to-face service and human interaction become luxury items offered solely to high-value customers, as is already happening in many automation-focused banks? Turner agrees that:

it is inconvenient to have to go to the robot. But it’s not ‘one size fits all’ for us. There is a trade off between the attractions of human and time-slot-based deliveries.

So what of the robots themselves? The machines are AI-enabled, which means over time they learn about the local terrain and map their environments, so their performance and knowledge will improve.

But physically, the robots are knee-high to an adult. This would seem to present a pedestrian hazard in busy streets – particularly where blind or partially sighted people, shoppers coming out of doors with bags, running children, elderly or disabled people (not to mention anyone with prams, or on bikes, skateboards, or mobility scooters) might be at risk of colliding with them and injuring themselves.

Turner counters that 10,000 miles of testing to date in busy streets have presented no such problems. That’s great and you can’t argue with that, but it’s conceivable that confirmation bias has crept into the programme - no company would set out to put vulnerable pedestrians at risk in crowded streets to see if any of them might be injured. This may present problems in a real-world implementation.

Turner explains that Starship has considered the visibility problem by fitting a flag to the machine, which features eye-level, flashing LEDs. He adds:

This isn’t the finished design. There are a number of iterations of the robot. We’re currently on prototype E at the moment. We’re making improvements to the battery life, the drive train, and the onboard sensors.

And is the robot secure from the new forms of crime that might emerge in a world of autonomous devices, such as theft from (or of) delivery robots? The robot is alarmed, GPS located, sensor-packed, steel reinforced, and contains an onboard camera, says Turner, and the end-customer of each delivery will receive a one-time code on their phone to open the lock.

My take

Pedestrians aside, it stands to reason that the main obstacles to autonomous, AI-enabled robots and self-driving vehicles on our roads and pavements (and in our skies) are legislative. The technologies exist, together with the business will to implement them – in many cases to save money, but in others to satisfy new or emerging demands.

But let’s hope that common sense is still delivered to our doors in the process. As MIT Media Lab’s Joichi Ito said at the World Economic Forum in Davos this year, one of the problems with the development of AI isn’t the technologies themselves, but sometimes the binary thinking and narrow, diversity-lacking world views of their programmers.

As a result, no one asks the questions that are obvious to outsiders. But once autonomous machines are on the streets, we are all outsiders. At that point, we can only hope that the messy human world has been anticipated – and experienced first hand – by programmers.

Our romance with technology should be more than a simple on/off affair, more than a simple yes/no answer. This is why both vendors and their clients need to consider the worst that might happen very seriously, while keeping the customer satisfied. Because if they don’t, a highly-paid lawyer certainly will.