Edge computing is a natural enabler for distributed robotic applications that need low latency and adaptability, yet current approaches offload computation only to edge nodes and ignore cases where onboard processing would already meet the performance targets. The paper proposes a framework that merges the offloading decision and resource orchestration into a single algorithm, comparing allocation strategies against accuracy, latency and energy goals. Results show the approach admits more robotic functions while still meeting strict performance targets, delivering substantial gains in overall system utility.