Ray tracing produces information regarding the precise rays between a transmitter and a receiver and is thus suitable in point-to-point scenarios where there are only a few receiver locations of interest. However, if the number of prediction locations is considerably large, ray tracing will suffer from a long running time, especially in scenarios where there are many complex obstacles which will incur more data-intensive operations such as intersection tests. Besides, ray tracing treats the calculation of neighbour pixels equally the same i.e., the computational time is roughly linearly proportional to the number of receiver locations.
In contrast, ray launching computes the rays from the emitter. Ray launching is an image-sampling method which uses discrete rays by an angle. Inevitably, gaps will be created gradually after the rays undergo reflections and diffractions. In order to solve this problem, a reception sphere can be used to capture the missing rays. More rays can be launched to improve the accuracy but this will slow down the computation. In general, ray launching walks through the rays and computes the reflections, transmissions and diffractions iteratively. The pixels gain experiences from its previous pixels along the same path, which is faster than ray tracing. Ray launching is suitable in point-to-many scenarios, such as coverage prediction. The ray launching may be more suitable for wireless network planning and optimisation in indoor scenarios because generally it is computationally more efficient than ray tracing and it provides a relevantly acceptable level of accuracy.
