We consider a network of smart sensors for edge computing application that sample a signal of interest and send updates to a base station for remote global monitoring. Sensors are equipped with sensing and compute, and can either send raw data or process them on-board before transmission. Limited hardware resources at the edge generate a fundamental latency-accuracy trade-off: raw measurements are inaccurate but timely, whereas accurate processed updates are available after computational delay. Also, if sensor on-board processing entails data compression, latency caused by wireless communication might be higher for raw measurements. Hence, one needs to decide when sensors should transmit raw measurements or rely on local processing to maximize overall network performance. To tackle this sensing design problem, we model an estimation-theoretic optimization framework that embeds computation and communication delays, and propose a Reinforcement Learning-based approach to dynamically allocate computational resources at each sensor. Effectiveness of our proposed approach is validated through numerical simulations with case studies motivated by the Internet of Drones and self-driving vehicles.