Autonomous mobile robots bring a different hazard: navigation, not just contact
Fixed industrial arms and cobots operate within a bounded reachable workspace, which is part of what makes their risk assessment tractable, the set of places the robot could possibly be is limited and known. Autonomous mobile robots (AMRs) and warehouse robots break that assumption, they move through shared space that also contains people, forklifts, and changing obstacles, which introduces navigation-related hazards that a fixed arm never has to deal with: will the robot correctly detect and yield to a person stepping into an aisle, will it behave predictably at a blind corner, will it stop appropriately if its sensors are degraded by poor lighting or a dirty lens. Standards addressing this space, such as ANSI/RIA R15.08 for industrial mobile robots, extend safety thinking from 'is contact force survivable' toward 'does the robot reliably perceive and appropriately react to people and obstacles while it's in motion around a shared facility.'
This is a genuinely different engineering problem from force limiting a robot arm, it depends heavily on perception reliability, sensor coverage, and predictable behavior under uncertainty, rather than primarily on mechanical force characteristics. A mobile robot's safety case has to account for things like sensor blind spots, degraded performance in poor conditions, and how the robot behaves when it's genuinely uncertain what it's looking at, questions that don't really have a fixed-arm equivalent.