The automotive industry has already been forced to solve this class of problems. While full autonomy is still being developed and validated, autonomous and connected vehicle platforms have long been required to run complex perception and planning workloads under strict safety, real-time, and regulatory constraints. What emerged was not a single autonomy solution, but an execution architecture, running mixed-criticality systems that allow high-performance AI and deterministic, safety-controlled execution to coexist on shared hardware. That same pressure is now hitting autonomous mobile robots (AMRs) and connected robotic systems.
In production, automotive platforms paired with QNX OS and or hypervisor-based isolation have become a driving foundation. QNX allows for real-time systems, fault containment, and safety integrity. Hardware-accelerated perception runs at scale. When this architecture is applied to robotics, something important changes.
Safety stops being a bolt-on emergency brake. Instead, it becomes a continuous supervisory layer embedded directly in the motion pipeline. Autonomy can propose behaviors freely, but execution is mediated through safety-controlled pathways that constrain speed, direction, and separation in real time. Rather than just defaulting to binary stops, robots move in a state of controlled motion - slowing, biasing trajectories, or proceeding confidently based on verified conditions.
Infrastructure-mounted perception systems, built on the same hardware and software foundations as those in vehicles, can see beyond a robot’s local field of view. Blind corners, occlusions, and congestion are detected using identical sensing and acceleration pipelines, then shared as validated situational awareness rather than direct control. Motion authority stays with the robot, but its safety domain now reasons over a wider world.
However, what’s emerging isn’t an “automotive solution for robotics,” but a shared architectural foundation.
Automotive forced the maturation of mixed-criticality compute, real-time operating systems, and cooperative perception under extreme constraints. Robotics is now in a position to reuse those patterns - reducing risk while enabling higher autonomy, tighter human interaction, and scalable deployment.
What’s missing is a clear, practical reference for how these architectures translate from AV/CVs to AMRS and external mounted perception systems - what changes, what doesn’t, and where the real safety boundaries move when execution becomes a continuous, supervised process rather than a binary stop condition. That gap matters for anyone designing next-generation robotic platforms meant to operate at scale, around people, and across shared infrastructure.
Interested in a deeper exploration of this architectural convergence? Get in touch.
About the author
Winston Leung is Senior Manager at QNX.