Outsight’s Software Solution Powers Automation Across BMW Factories

Frequently Asked Questions

• How does infrastructure-based LiDAR guide autonomous vehicles inside a factory compared to onboard sensors?

  Infrastructure-based LiDAR mounts sensors on fixed points throughout a facility, such as ceilings, poles, and gantries, giving the system a complete, occlusion-free view of every vehicle in motion across the entire plant floor. An onboard sensor sees only what is ahead of one vehicle and cannot track other vehicles simultaneously. The infrastructure model centralizes the perception workload: one software stack handles all vehicles at once, which simplifies safety certification and allows the autonomous-driving controller to receive a shared, site-wide picture rather than each vehicle reasoning in isolation. Outsight applies exactly this approach inside BMW factories through its Roadside Autonomous Driving solution, where infrastructure-mounted LiDAR feeds a unified 3D understanding of the production environment to guide autonomous vehicles without placing any perception hardware on the vehicles themselves.

• What safety certification does BMW's in-plant autonomous vehicle system carry?

  The Automated Vehicle Marshalling system deployed at BMW's Dingolfing plant has obtained CE certification, the conformity marking required for machinery and automated systems operated in European industrial environments. CE certification verifies that the system meets EU safety, health, and environmental requirements. This certification was achieved through Outsight's RAD (Roadside Autonomous Driving) solution, which uses infrastructure-based LiDAR perception to guide autonomous vehicles through factory environments without relying on onboard sensors. The certified system then enabled rollout planning across additional BMW facilities in Leipzig, Regensburg, Oxford, and the planned Debrecen plant in Hungary.

• How many data points per second does the BMW factory LiDAR perception system process?

  The Roadside Autonomous Driving solution deployed at BMW's Dingolfing plant processes up to 600 million LiDAR data points per second across the full system, with the per-vehicle guidance workload running at 60 million points per second. This throughput is delivered through Outsight's infrastructure-based approach, where LiDAR sensors are mounted in the facility rather than on each vehicle. That capacity is sufficient to guide approximately 300 vehicles per day through routes of more than one kilometer from assembly halls to finishing zones, without human intervention.

• What role does each partner play in the BMW Automated Vehicle Marshalling system?

  Three companies provide distinct layers. Hesai supplies the physical LiDAR sensors mounted throughout BMW's facilities. Outsight's software ingests the raw sensor data and converts it into a real-time 3D perception stream, tracking each vehicle's position and trajectory through the SHIFT platform, which processes data end-to-end in under 50 milliseconds. Embotech provides the autonomous driving stack that consumes that perception stream and issues motion commands to the vehicles. The three components are independently replaceable because Outsight's software is hardware-agnostic across more than 210 LiDAR sensor models, reflecting the multi-vendor compatibility that underpins its Infrastructure-based Physical AI approach.

• Can the same LiDAR perception software used for factory vehicle guidance also handle worker safety?

  Yes. Infrastructure-based LiDAR perception classifies both vehicles and people within the same 3D scene, distinguishing cars, trucks, pedestrians, and equipment simultaneously. In an automotive plant, this means the system can detect a worker entering a vehicle path and trigger a stop command before a collision occurs. Outsight's SHIFT platform applies exactly this principle across BMW factories, where the same classification pipeline that tracks autonomous vehicle trajectories also assigns bounding boxes and behavior labels to people. Worker-safety monitoring and vehicle guidance become complementary outputs of one shared sensor network rather than separate systems, reducing infrastructure overhead while covering both operational and safety requirements from a single deployment.

• Why is a hardware-agnostic perception stack important for a multi-plant factory rollout?

  Large manufacturers rarely standardize on a single LiDAR sensor model across all facilities. Ceiling heights, bay widths, outdoor versus indoor exposure, and local procurement contracts all push different plants toward different sensor hardware. A software stack tied to one sensor vendor forces a costly hardware standardization project before deployment can scale. A hardware-agnostic stack, compatible with sensors from multiple manufacturers, allows each plant to select the sensor that fits its physical layout and budget while running the same software and producing the same structured perception output. Outsight's SHIFT platform is built on exactly this principle, supporting hardware from Hesai, RoboSense, Ouster, Velodyne, and others, which is part of why BMW has deployed it across multiple factories. That multi-vendor compatibility simplifies IT integration and operator training across a global production network without forcing a disruptive sensor consolidation program first.