In a recent Monday in February the usually bustling streets of Beijing’s Shunyi District were nearly empty. As a trickle of masked delivery persons bicycled by long lines of dusty parked cars, a four-wheeled, yellow-and-black contraption came whirring through — one of many AI-powered autonomous vehicles that have been navigating Chinese streets in recent weeks in response to the COVID-19 outbreak.

The vehicle, Modai (“Magic Bag”), was delivering much-needed groceries to communities in the district. Modai was developed and modified for grocery delivery by Chinese O2O local life service company Meituan.

Running slightly faster than bicycles and on non-motorized lanes, Modai is around 156 cm tall and weighs 500 kg. Stickers on the vehicle proclaim “Non-contact, Safe delivery” and “Disinfected today.” A Modai can drive about 100 kilometres without a charge and carry up to 100 kg of goods, enabling it to make several deliveries per trip.

Previous Challenges

Autonomous delivery is not new in China. JD.com’s self-driving delivery vehicle, the first to be domestically developed, began road testing in September 2016. Meituan launched its W Project team focusing on the R&D of self-driving delivery cars the following month, and was soon followed by companies like Cainiao, Wise Walker, Xingshen Intelligent Technology, Zhong Yun Intelligent Machinery, and Yiqing Innovation.

However, deploying such delivery vehicles can hardly be profitable unless it’s done in areas with many end users — if the number of users is too small, it will be difficult for companies to cover both the R&D and operating costs.

The cost of a small self-driving vehicle is around CN￥100,000 (~ US$14,067) on average while a medium-sized model is about CN￥200,000 (~ US$28,134). The average vehicle distribution cost is around CN￥7-8 ( ~US$1) per order, according to Huaxia Xia, general manager of Meituan’s self-driving delivery department.

Challenges facing autonomous delivery companies include optical components meeting production capacity and efficiency requirements and scaling operations. In addition, the current laws in China haven’t yet specified whether self-driving delivery vehicles should be categorized as “robots” or “motor vehicles.” Existing regulations do not permit self-driving vehicles to legally run on public roads, but robots are allowed to do so.

Technological bottlenecks and grey areas in laws and regulations combined with China’s dense population and complex traffic conditions have restricted large-scale testing and operations of self-driving delivery vehicles. Most previous real-world applications were limited to relatively closed scenarios such as schools, gated communities, parks, etc.

New Testing Grounds

The outbreak of COVID-19 however changed the game. Suddenly many people were restricted to their homes and had to rely on delivery services for food and other daily necessities. Meituan’s service stations became busier than ever, with human drivers delivering vegetables, fruits, rice, noodles, oil, etc. The escalating demand was a serious concern. The two stations in Shunyi for example have a total coverage area of 86 surrounding communities with 100,000 potential users, said Da Li, product head at Meituan’s self-driving delivery centre

Self-driving delivery vehicles like Modai seemed the perfect solution to the COVID-triggered surge of online orders for grocery shopping and the need to reduce interpersonal contact to slow disease spread. Because Shunyi District streets were almost empty of regular traffic, they were turned into a massive testing ground.

With a green light to test Modai on public roads, the Meituan team spent two days modifying the vehicles so they could carry non-standard and often heavy loads of fruits and vegetables, and coming up with a simplified method for retrieving goods from vehicles. To design effective delivery plans, the team also needed to consider in advance the specific situations in each community and pinpoint the best delivery locations.

To reduce delivery dangers and interruptions, the team took weather conditions into consideration. Low visibility on rainy days for instance put high demands on Modai’s sensors, while network signal interruptions could cause significant delays in the scheduling system. The CV systems were adjusted to work under more scenes.

Meituan debuted its fearless Modai on February 10 on publicly tested roads. Departing from Meituan’s Shunyi service station, the car actively avoided obstacles and successfully identified traffic lights and other road signs before adroitly parking at the receiver’s door.

The whole delivery process is conducted under the supervision of security officers who check the status of cars through a remote monitoring video and call residents when cars arrive to ensure smooth delivery completion.

Da Li says that establishing cost-effectiveness of autonomous grocery distribution will enable the model to be applied to delivering restaurant take-away, medicines, and so on; and that large-scale deployment can solve transportation capacity problems and reduce costs.

Novel Solutions

The pressing need and empty streets have drawn a number of companies to autonomous delivery. In February, JD Logistics developed self-driving delivery vehicles in Wuhan and for the first time successfully delivered medical supplies to the Wuhan Ninth Hospital. The Suning Logistics 5G Wolong self-driving car also delivered its first order in Suzhou.

Shanghai-based Keenon Robotics meanwhile cooperated with Meituan and the China Hotel Association to build “non-contact restaurants” to provide food services and delivery while avoiding unnecessary direct contact and ensuring the safety of both customers and restaurant workers.

Keenon also launched a “smart solutions for the epidemic areas” plan, sending hundreds of delivery robots to nearly 100 hospitals in the affected areas across the country, such as Wuhan Union Hospital and Wuhan’s temporary mobile cabin hospital. The robots helped transport medical supplies, deliver meals for doctors and patients, and complete other emergency tasks as needed. Use of the robots not only avoids cross-infection but also reduces the workload of medical staff.

Established in 2010, Keenon focuses on R&D and production of indoor unmanned delivery robots in a variety of sectors. The company’s delivery robots are active in over 400 cities across China.

In autonomous food and medicine delivery and beyond, the increasingly mature AI industry is actively engaged in countering the COVID-19 outbreak, by accelerating early-stage research, assisting with front-line diagnosis and on-site prevention and control, establishing back-end information platforms, and offering practical solutions for affected communities.