Radar sensor networks

Abstract

This work aims at interference mitigation in recently booming automotive radar networks, via careful beam forming techniques based on Geometric Algorithms and Optimizations. We first thoroughly explore a relatively simple 2-dimensional case, where the environment is completely static and the ego sensor node has full knowledge. After developing a Dynamic Programming algorithm for the unity gain scenario and further incorporating a simpler gain model via introducing dummy interferers, we simulate the algorithm for different environment instances and analyze how the performance and optimal number of target detections vary. Then we present multiple other directions in which this problem can be extended sensibly. First attempt was to allow beam overlaps, but its premise had practically limitations. Then we tried to derive an approximation factor for our DP solution in 2D, and also formulated a Mixed Integer Program for the 3D case. But in both of them, a major hindrance to algorithm design was the complex objective function, which needed to be addressed to proceed further. And finally for the3D sensing region scenario, we try to leverage our 2D solutions by Stereographically projecting the points on the surface of the sphere to a 2D plane. Even though the mapping didn't lead to the exact 2D problem we had handled, we ended up with the analytical expression for resultant objects in the 2D plane after projection, using which we can try to arrive at a problem definition almost equivalent to set cover in 2D..