On 23 July, Apollo Go received Hong Kong’s first driverless test licence, and on 27 July it began public-road testing on the airport island with no safety driver inside, the world’s first fully driverless project in a right-hand-traffic system. One day later its sixth-generation RT6 appeared on the streets of London, where on 28 July it launched public-road tests with Uber and Lyft’s Freenow, becoming the first Chinese Robotaxi firm in London’s public roads.
Two global ride platforms choosing Apollo Go together signals its capability moving from self-validation to the market test of an international mobility ecosystem. From Hong Kong to London, nearly 10,000 kilometres apart but one day apart in time, a clear Robotaxi globalisation path appeared.
Hong Kong proved the right-hand fully driverless breakthrough; London then takes the next baton, testing whether the capability can enter one of the world’s most representative urban transport systems. Apollo Go now covers 27 cities worldwide and, on 29 July, Slovakia’s president rode the RT6 and praised its maturity.
Why right-hand markets lagged
The global Robotaxi map now has a clear divide. China and the United States, with large home markets and dense road data, led the shift from closed tests to public roads to driverless scale. Right-hand-traffic markets, from Britain, Japan, Australia and Singapore to Hong Kong, share left-side driving but differ in road structure, law and driving culture, raising local adaptation and regulatory cost.
The decisive gap is data. Most complex scenarios the leaders have seen came from left-hand-traffic roads; right-hand environments offer far less training and validation data, and almost no experience of the driverless stage. Yet right-hand markets are not niche: more than 70 countries and regions drive on the left, covering over 2 billion people, including high-value cities like London, Tokyo, Singapore and Sydney.
Mirror is not enough
Moving from left to right traffic changes more than the wheel side. Roundabouts, conflict points, crossing flows and pickup positions all flip, and the prediction and planning layer must relearn right of way. Mirroring data helps efficiency, but real traffic is not symmetric. Singapore drives on the left yet follows a “yield to the right” rule like left-driving countries, a case mirroring cannot brute-force.
Apollo Go’s Hong Kong run was a three-layer validation: adapting road rules, migrating driverless capability, and closing the regulatory and safety loop. London then upgrades the test to generalisation under narrower roads, denser traffic and stronger non-standard interaction. There, the system must judge not just whether a target moves but whether it will enter the path, modelling interactions among multiple actors.
The Hong Kong and London runs form a progressive chain. Hong Kong proved driverless can cross the left-right line; London tests how efficiently the same base generalises to a harder city. In global Robotaxi competition, migrating a mature system to more traffic systems at lower cost may decide commercial viability.
Editor’s note: This is an adapted translation of the original CheDongXi report. It has been trimmed and restructured for readability for an international business audience.