At the beginning of September 2025, Swiss Post and Migros Online unveiled a trial in Regensdorf, near Zurich, to deliver parcels directly to recipients’ doors using a quadruped robot developed by Rivr. Swiss Post will act as logistics provider, while Migros Online’s deliveries will be integrated into existing channels. The goal is to assess how automation can improve efficiency, ease physical strain on staff and enhance customer experience without disrupting established logistics processes. Human staff will supervise the entire process to ensure safety. The initiative also seeks to evaluate whether robotic solutions can reduce walking distances and heavy loads for workers, while supporting sustainability in service delivery.
The collaboration with Swiss Post is not Rivr’s first field test. The company has already run pilots in various urban settings, from the United States to Europe, where robots have faced challenges such as crowded pavements, pedestrian crossings and narrow access points. On US university campuses, for instance, the machines managed deliveries in mixed-use areas, feeding a central data hub with information to refine artificial intelligence. In Europe, additional trials took place in older neighbourhoods with outdated infrastructure and uneven routes, testing adaptive navigation and mechanical endurance.
More recently, Rivr has partnered with Just Eat Takeaway.com, one of the world’s leading food delivery platforms. Since August 2025, Rivr robots have been in service in Zurich, carrying out live restaurant deliveries with plans to expand to other European cities before year-end. Potential future uses include retail and convenience store deliveries. For this service, the robots are configured for high-frequency, fast-response missions. They carry meals in a 40-litre insulated compartment with internal dividers, ensuring protection and stability during transport.
With a top speed of around 15 kilometres per hour, the robots can operate in adverse weather such as rain, snow or strong winds, providing continuity of service. The delivery process mirrors traditional models: the restaurant places the order inside the sealed box, the customer receives a notification upon arrival and unlocks the compartment to retrieve their meal. All units are monitored in real time from an operations centre, which can intervene immediately if required. According to Rivr CEO Marko Bjelonic, the partnership offers a glimpse of how automation could become seamlessly integrated into cities, making deliveries more intuitive, safer and more citizen-friendly.
At the heart of Rivr’s platform is a General Physical AI, an artificial intelligence designed to coordinate large numbers of robots in real-world environments among people. Training combines reinforcement learning in advanced GPU simulations, allowing robots to develop mobility and manipulation skills within days, and supervised learning based on real-world data. Human experts remotely guide robots through complex situations, enriching the dataset and refining their behaviour. This generates a virtuous cycle known as the Data Flywheel: every delivery produces fresh information, which is processed centrally and redistributed to the fleet, enhancing autonomy and progressively reducing the need for human intervention.
From an engineering perspective, Rivr has opted for a hybrid platform combining wheels and legs. Wheels deliver speed and range, up to five times greater than leg-only systems, while legs enable navigation of stairs, kerbs and uneven terrain. Advanced sensors including cameras, LiDAR and multimodal perception systems allow for safe environmental recognition and navigation. The robots are also equipped with a robotic arm capable of opening gates or placing parcels at doorsteps, as well as an insulated container designed for groceries and medium-sized packages. Central electronics housed in the torso coordinate movement and data processing.
The flagship model, Rivr One, has been developed to provide an integrated and flexible delivery service, capable of operating across both the last mile and the last hundred metres. Its design allows it to function autonomously on streets or in direct interaction with autonomous vans, collecting parcels for final delivery to the customer.
Rivr highlights the robot’s performance features, including walking speeds faster than a human pace, emergency stop systems both automatic and manual, and improved visibility with lights and indicators for safe operation at night. The design ensures safe interaction with pedestrians, cyclists and even pets, while avoiding sensitive areas such as gardens and green spaces. High energy autonomy and rapid recharging support continuity of service, while payload capacity and the option of unattended deliveries extend its range of applications.
Equally significant is the management infrastructure behind the fleet. All robots are connected to a cloud-based system that enables continuous mission oversight and remote operator intervention when needed. This hybrid approach, blending autonomy with supervision, is considered crucial for safety and industrial scalability. Every mission is not only a logistics operation but also a training opportunity that makes the entire system more efficient.
