John Wall, President at QNX, a division of BlackBerry.
getty
For decades, robots operated behind fences in controlled environments, physically separated from people and optimized for predictability. Safety was achieved through isolation: Defined zones, fixed layouts and simple overrides stopped motion whenever uncertainty appeared. That model worked because human interaction was the exception, not the rule.
Today, that assumption is fading fast.
As robots step into shared human environments—think sidewalks, hospital corridors, retail spaces, construction sites and more—the safety stakes rise and the margin for error narrows. Conditions change constantly in ways no set of rules can fully anticipate. Industry research from QNX’s Inside the Robot: Architecture Benchmark Report highlights a growing gap between ambition and readiness, with only a quarter of robotics developers confident their current architectures can scale significantly to meet future demands.
Robots are arriving in shared spaces faster than the architectures that support them can hold up.
Many organizations still approach robotics with a legacy mindset, assuming that precision engineering and predefined behaviors are enough to ensure reliability. The defining challenge facing the industry is no longer what robots can do, but whether their underlying architectures are designed to manage this uncertainty continuously without defaulting to constant stops that undermine trust.
Deployment is accelerating, but many robotic systems are scaling on software architectures seldom designed for continuous uncertainty. When perception workloads spike, sensors degrade or timing becomes inconsistent, most systems lack the guarantees needed to respond predictably. Instead of adaptation, they fall back to their only reliable option: stopping.
The result is mismatch between expectations and reality. Leaders expect robots to operate like dependable teammates in human environments, but the underlying architecture treats safety as an external override rather than a continuous constraint on motion. The robot stops, availability collapses and trust erodes. This isn’t because the robot lacks intelligence, but because the system was never designed to manage uncertainty in motion, only to revoke it.
In controlled environments, safety could be validated upfront. Boundaries were fixed, behaviors were known and stopping was acceptable. In shared environments, that model changes and safety can no longer be something proven once; it must be enforced continuously, under changing conditions.
Consider a delivery robot navigating a crowded city sidewalk. Traditional designs rely on predefined speed limits and emergency-stop thresholds. Dynamic, software-driven safety works differently. As pedestrian density increases or sensor confidence drops, the system continuously constrains motion: reducing speed based on verified stopping distance, biasing trajectories away from people, limiting acceleration and narrowing feasible paths to those that can be safely executed.
Safety isn’t a single intervention. It becomes an ongoing negotiation between perception, planning and control.
This approach demands a different architectural mindset. Safety logic must run with deterministic timing, independent authority and access to safety-relevant abstractions like clearance margins, confidence levels and feasible trajectories.
For leaders, this marks a fundamental shift in responsibility. Safety is no longer a box checked at deployment but, rather, a property of how software is structured, scheduled, supervised and evolved over time. Organizations that continue to treat safety as an external layer will deploy robots that are technically compliant, but operationally brittle. Those that rethink safety as a continuous, architectural property will be the ones whose robots remain both safe and dependable in the environments they are increasingly expected to share with people.
As robotic architectures become more software-defined and more connected, the boundary between digital failure and physical harm is also disappearing.
Connectivity enables scale—fleet coordination, remote diagnostics, OTA updates—but it also expands the system’s attack vectors. It’s no coincidence that more than half of robotic developers in QNX’s same study now cite cybersecurity requirements as one of their most difficult regulatory hurdles.
In robotics, a compromised system doesn’t just leak data—it moves. Manipulated sensor inputs, delayed or corrupted control messages, or unauthorized access to processes can directly alter motion, timing and behavior.
What makes cybersecurity uniquely critical is that trust must be continuously enforced inside the robot itself. Systems must remain safe not only when components fail naturally, but when data is corrupted, communication is interrupted or execution is violated. That requires architectures that assume interference will occur and is designed to contain it through isolation and deterministic execution such as those provided by a microkernel OS.
As robotics systems grow to be more software-defined, many teams are discovering that proving a system is safe, secure and compliant now takes as much effort as building it. Regulatory review timelines stretch, validation cycles repeat and certification becomes a bottleneck.
Today’s deployments look fundamentally different. Robots operate continuously alongside people, where human presence is normal and stopping is no longer the default response. Instead of halting at the first sign of uncertainty, robots are expected to slow, yield, reroute and adapt while remaining operational.
This shift changes what the safety case must prove. It’s no longer enough to show that a system stops when something goes wrong. Modern safety arguments must demonstrate that the robot can manage uncertainty in real time.
As a result, integration, validation and certification are becoming constraints because they now reach deeper into the software stack. Platforms built on proven safety and security foundations pre-certified to industry standards can reduce this burden.
Moving robots into everyday environments is one of the most significant transitions the industry has faced, and one of the most unforgiving. Leaders who prioritize software that supports deterministic performance and continuous safety validation will have a much easier time leading robots beyond the fences and staying the course.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。