Momenta has begun running driverless Robovan parcel deliveries in Xiangcheng district of Suzhou, in Jiangsu province, east China, including overnight routes. The vehicles run on the company’s R7 world model and take a map-free approach that removes any dependency on high-definition (HD) maps.

Dropping HD maps is the commercially significant part. Surveying and maintaining them is slow and costly, so a map-free stack can reach new districts far faster, which is how Momenta expects to bring per-delivery costs down. Cost per drop is the metric that decides whether this scales.

The claim rests on data drawn from passenger cars. Momenta says production fleets running its driver assistance software have covered more than 12bn km, distilled into over 100m curated data segments, and that the learning carries across to delivery vans.

That transfer is the open question. Loading bays, kerbside stops and depot yards are not where a passenger car fleet spends its kilometres, and Momenta has not said how many Robovans are running, on how many routes, or when the deployment scales. Night running is the easier case, with clear roads and fixed loads.

The firm wants a single foundational world model spanning passenger cars, Robovan, robotaxi and robotruck, and says it aims to double logistics and mobility efficiency within 10 years. Momenta also supplies the end-to-end driver assistance in the BMW X5 and iX5 for China, which gives some measure of how far its software has already travelled.

Source: Momenta