Field robots are already at work in areas such as mining, agriculture, and environmental monitoring. Future applications will require larger-scale systems with multiple robots that can work autonomously in difficult conditions that push the limits of sensing, perception, planning, and communication. This talk will focus on coordinated object perception in outdoor environments and discuss how occlusions and sensor characteristics limit detection and classification accuracy, how to reduce uncertainty by combining observations from carefully chosen viewpoints, and how motion planning and mobility can play a major role in addressing these challenges through an active perception approach. The talk will give an overview of our research in developing new methods for active perception in decentralised teams that integrates ideas from perception and planning, information theory, and game AI to allow robots to coordinate their behaviour in real time. We have implemented our methods in physical systems with wireless communication, and the talk will describe several examples of multi-robot systems that search realistic environments to detect and classify objects. The talk will also look towards future systems and discuss key open challenges at the frontier of mobility, sensing, and communication.