Validation of hybrid systems is complex due to interactions of both continuous and discrete dynamics. Simulation is the most widely used form of system validation using a combination of random and constrained-random tests. Directed tests are promising since orders-of-magnitude less number of directed tests can achieve the same coverage goal compared to random tests. While directed test generation is well studied for digital designs, it is still in its infancy for hybrid systems. In this paper, we propose a method for automatically generating directed tests for hybrid systems. The test generation scheme is based on the Rapidly Exploring Random Tree (RRT) algorithm. In contrast to existing methods of using RRT for validation, that tries to reach targets (functional scenarios) from the initial state, we propose to employ reverse RRT that starts from a target and tries to reach the initial state. This enables us to generate an accurate testcase for both functional scenarios and interesting corner cases. We also demonstrate how learning can drastically reduce the test generation time between a set of similar test generation instances. Our test generation algorithm is upto 33 times faster (average 10 times) compared to state-of-the-art forward RRT techniques. Additional test time reduction by a factor of 1.7 (average 1.27) can be achieved by exploiting learning based testcase generation.