Unmanned Systems Technology 015 | Martin UAV V-Bat | William Sachiti | Sonar Systems | USVs | Desert Aircraft DA150 EFI | SeaCat AUV/ROV | Gimbals

19 challenges in the UK. “If I drive a US car on UK roads I will struggle because I’m on the ‘other’ side of the road, with different traffic priorities. So whenever someone drives off an A road, that’s when our AI kicks in, when there are no more road markings. It’s trained with images, and the data is annotated.” The Academy used a chassis from Renault to build the first prototype of the Kar-go, with battery packs also from Renault that are the same as those from Nissan for the Leaf electric car. “We used the Twizy chassis to build our first prototype, but it needed a lot of hacking about so we were not happy with it,” Sachiti says. “However, we found an engineer who had retired from [racecar developer] McLaren, who designed a chassis for us, so we probably have the first new chassis designed for driverless cars. “There were a few things we wanted from the design, such as the ability to swap batteries, no wing mirrors and removing the bits you find for humans such as reinforcement in various places.” Regulations still require an operator to be present though, which presented a problem for a vehicle designed not to have one. “So we said, ‘Let’s design this for two years ahead for when the legislation changes.’” At the moment the Kar-go can be operated remotely, but the aim is to have a driverless delivery vehicle that travels from a local distribution depot to the kerb outside a house or block of flats. That means travelling on dual carriageways and smaller roads at speed. Customer collection The vehicle messages the recipient, who comes to the curb to collect the package. A conveyor system inside the vehicle then delivers the right parcel to the recipient (via a cloud-based management system). This is different from other hybrid systems that combine vans and smaller, pavement- based systems. “In my opinion there are too many layers when you have a van and then send a small robot to the house. It’s cheaper and there are fewer break points if you go straight from the distribution centre – the same Kar-go will go from the centre to your house. “It has a range of 50 miles and carries 24 packages. We send you a link to where the vehicle is, and when it’s outside it automatically calls you. The most important thing is that this is on- demand delivery – it will only come to where you are.” In this way the use case determines the technology requirements. The vehicle has to travel on a variety of roads but it cannot be used on the pavement. Sachiti’s strategy has been to team up with suppliers to create the system, from batteries to comms systems that connect to the cloud and to customers. “We are leasing the battery from Renault [which is still the Leaf system], who do it better than us,” Sachiti says. “The same goes for the electric motor, and we are still exploring what the best options are for that.” The company has developed its own hybrid image sensors (which it is patenting) that are optimised for unmarked roads. “We have a really good way of detecting objects such as paper bags, cups – things you get on residential roads – so we built a hybrid sensor,” Sachiti says. “Our team believes that Lidar should be for redundancy, and that cameras should be sufficient.” Another partner is specialist William Sachiti | In conversation Unmanned Systems Technology | August/September 2017 The Kar-go stops by the kerb and delivers the required package using an internal conveyor system