Abstract

Edge AI is reshaping industrial robotics by moving perception, inference, and coordination closer to sensors, actuators, and motion controllers operating under strict timing constraints. This chapter examines latency optimization in real-time robotic control as an architectural and benchmarking problem rather than a hardware-selection issue alone. It develops a layered reference architecture linking sensing, middleware, inference runtimes, control execution, safety envelopes, and edge-cloud orchestration. It also proposes a benchmarking framework centered on end-to-end latency, jitter, deadline misses, control-age, energy efficiency, and thermal stability. Drawing on literature in edge computing, ROS 2, embedded AI, deterministic networking, and industrial robotics, the chapter synthesizes design trade-offs and presents literature-grounded industrial case studies in machine tending, vision-guided handling, and mobile robotics, concluding with implementation guidance and a research agenda.