Autonomous Vehicle For Last Mile Deliveries

| Transport

The Kar-Go pod is bringing autonomy to the streets of suburban Surrey

Trials begin in Surrey for making contactless deliveries from local businesses to suburban homes

A successful initial testing phase of the Kar-go pod along fixed routes in London has led to autonomous vehicle company, Academy of Robotics, extending its test to the next phase serving residential addresses in Surrey.

The distinctive looking Kar-go is an electric vehicle, which was designed by Academy of Robotics to reduce the cost of ‘last-mile’ deliveries and cut harmful emissions. Its operating system has been designed to navigate unmarked roads outside city centres so it can make deliveries not just in urban environments, but also in more rural and suburban areas.

Founded as a start-up out of the University of Aberystwyth in Wales by William Sachiti, the company pulled together the expertise of autonomous motoring algorithm specialists and vehicle manufacturer, Pilgrim Motorsport to package the Kar-Go pod.

Following years of data gathering and testing with a combination of simulators and live technology trials in real-world environments, the team focused on the complex arrival and departure aspects such as where the vehicle can pull up and park.

The result is a vehicle that uses route information taken from route-learning vehicles along with cameras, sensors and sophisticated artificial intelligence to navigate autonomously to pre-arranged destinations.

Academy of Robotics made medical products the focus of Kar-go’s first deliveries in a proposal that was awarded funding from UK Research and Innovation to conduct trials in Hounslow during 2020. The trials involved the Kar-go delivery bot collecting medicines from pharmacies and delivering them to care homes and residential addresses.

Artificial Intelligence

The vehicle was able to perform these deliveries along pre-learned routes using modular driverless delivery technology. This combines Continuous-Time Recurrent Neural Networks (CTRNNs) with CNNs (Convolutional neural networks) and Long Short-Term Memory (LSTM), to create a highly evolved top-level controller system which can ‘learn’ from the past, ‘perceive’ its environment and make any necessary corrections.

Kar-go’s unique combination of its Navigational Freespace and PilotNet decision-making software means the vehicle can navigate safely in complex situations such as roundabouts and junctions even without GPS.

Also, unlike many autonomous vehicles, Kar-go can drive on un-marked roads making it perfect for solving the challenge of last mile deliveries in rural and suburban areas as well as urban environments.

The vehicle uses a terrain-training approach to become an expert in its particular route so it can then focus on any abnormalities of that route.

Surrey Trials

In late January, the design team began driving the Kar-go delivery bot in Surrey to scan the proposed delivery routes in preparation for using the vehicle to make semi-autonomous deliveries later in the first half of the year. The AI has been designed with multiple fail-safe layers and in line with current regulations for testing autonomous vehicles, there will be a safety driver in the vehicle at all times and the team will gradually increase the degree of autonomy used as the trials progress. In all cases, there will be the ability to control the vehicle remotely from a local command centre.

Commenting on the choice of Surrey for the latest trial, Sachiti said, “All too often new technology benefits only those living in cities and people living in more remote or suburban areas are left out. As a company founded in rural Aberystwyth, we were very conscious of this issue so we designed our AI to work where it could deliver the greatest benefit. So, the first areas we are rolling this out to are typical British suburban locations like Surrey.”

Cargo System

The patented package management system on the Kar-go includes the on-board storage and a customer mobile phone app.

On the car, there are storage compartments for different packages that can automatically select the package belonging to each customer. The hatch at the back of the vehicle is unlocked using a token on the recipient’s phone when the car arrives. The secure system then selects and releases this recipient’s package only, preventing further access to any of the other packages. As the vehicle moves on, this complex robotic system then re-orders the packages ready for the next delivery.

Being completely automatic, the system is contactless. Despite the car having a safety driver during the trial phase, there is no interaction between the driver and the parcel recipient. The role of the driver during the trial is only to be a minder to the autonomous control systems.

Safe to share the road with

The car’s design has the established UK automotive industry in its DNA and is manufactured by classic car experts, Pilgrim Motorsport under the supervision of former McLaren F1 designer, Paul Burgess.

To ensure the safe transport of its payload, the Kar-go not only has a driver in the initial trials but also the transmission allows for navigational movements to be conducted seamlessly by the onboard computer and the ultra-lightweight chassis is surrounded by bumper sensors and high-performance cameras, capable of bringing the vehicle to a halt if any object enters its designated safety proximity.

“In creating Kar-go, we wanted to take advantage of the strength of British motoring heritage and design talent and combine this with the latest AI technology to produce a vehicle which will deliver goods safely and sustainably,” Sachiti says.

As an entirely new vehicle type, the vehicle has been assessed by the Driver Vehicle Standards Agency (DVSA) and is now licensed to travel on the roads.

Related news

Read More News From Unspecified Company: