Simulation tools for wireless networks do not accurately capture the behavior of wireless protocols in real-world deployments. On the other hand, physical testbeds spanning large geographic areas are expensive to deploy, manage, and reconfigure. We propose to develop a miniaturized multi-hop wireless testbed, called Mint, that will provide a flexible and high-fidelity platform for protocol development, debugging, and testing, while significantly reducing the physical space requirement. A key architectural feature of Mint is its use of mobile robots to carry wireless network nodes, which can be programmatically controlled through a serial port based API. The space requirement is "miniaturized" by attenuating the radio signals in a controlled fashion. Additionally Mint will also support automated fault injection and analysis testing, remote testbed reconfigurability, and autonomic 24x7 operation. The Mint testbed will enable experimental investigation in a number of research projects, such as, cross layer optimization, optimal routing and transmission scheduling, effective routing cost metrics, resource management for multi-channel mesh networks, energy conservation for sensor networks, scalable geographic-based service provisioning, and modeling mobile low power wireless links. Our final goal is to enable researchers to remotely access Mint over the Internet in order to investigate their wireless protocols in a miniaturized setting.