In this paper, we show that the aggregate network throughput of a cognitive radio network can be significantly increased by using primary transmitter location sensing in addition to conventional spectrum sensing . To achieve this, we propose a novel multi-channel, cross-layer PHY-MAC protocol for CR ad hoc networks, which infers concurrent transmission opportunities from location sensing information. We identify four main problems and provide solutions which enable our location aware MAC. First, secondary nodes achieve location information without any interaction with primary nodes by combining RSSI and direction of arrival information (estimated from an antenna array). The location sensing capability can be integrated into the sensing module which also performs conventional channel sensing. Second, a distributed collaborative sensing algorithm is developed based on sensing module information which broadens nodes' knowledge of locations and spectrum occupancy. Third, a simple condition is proposed to control interference on primary network. Fourth, a decentralized scheduling scheme is presented to solve the MAC problem for location aware cognitive radio networks, in which channels are allocated to nodes based on their locations. Simulations show that the proposed approach significantly increases the performance of the CR network. The influence of protocol parameters on network performance is also investigated. ©2009 IEEE.