Analyze the role of edge computing in processing real-time data from autonomous vehicles.
Instruction: A sequential onboard pipeline has bounded worst-case delays of 30 ms sensing,4 ms queueing,12 ms inference and 4 ms command delivery under stated test assumptions. Its required deadline is 45 ms. Calculate the total, then evaluate a change reducing inference to 7 ms and identify what still needs validation.
For this exercise the stages do not overlap and the listed bounds are valid under the stated workload. Actuator response after command delivery is outside this defined deadline. Do not add unrelated percentile measurements as though they were deterministic bounds.
Updated
Worked answer
The original bound is 30+4+12+4=50 ms, exceeding the 45 ms deadline by 5 ms. Reducing inference to 7 ms makes the supplied sum 45 ms, exactly at the deadline with no allowance for unmodeled work or invalid assumptions.
Check that the optimization preserves required quality on relevant conditions and that conversion, memory, synchronization and scheduling costs are included. Validate the bounds under the target load, thermal behavior, faults and software version. Measured p 99 stage latencies are not generally deterministic bounds and cannot be summed to claim a p 99 end-to-end guarantee.
Onboard processing avoids an external network round trip, but it does not remove local sensing, queues, compute or command-delivery time. Cloud analytics may support asynchronous development; any remote dependency on this deadline path would need its own complete timing and outage model. Disconnected operation still depends on the validated architecture.
This is a defined compute-path budget, not a full stopping-distance or road-safety calculation. Physical actuator response and vehicle/environment dynamics need separate analysis if the system requirement is sensing-to-physical-response time.