Improved exit path identification with indoor USRP-based radar system

Radar is an object-detection system that uses radio waves to determine the range and direction of objects. It could be employed on an Unmanned Ground Vehicle (UGV) for autonomous navigation. Although the UGV could sense the surrounding environment by using optical sensors such as camera or infrared sensors, these sensors may not work well in some environments, such as an environment with low visibility or an environment with obstacles made of transparent or polished materials. In these cases, radar sensors could be employed along with these optical sensors to enhance the navigation capability of an UGV. To investigate this, a low-cost narrowband radar system based on USRP (Universal Software Radio Peripheral) hardware and C++/MATLAB programming is implemented to enable UGV autonomous indoor navigation. The main purpose is to investigate open doorway detection in an indoor environment utilizing this narrowband radar, which could be used for the UGV to detect the open doorway and navigate itself autonomously through the doorway. We first present the design and calibration of a power-based narrowband radar system. It was found that, when the radar faces the wall, the relation between the received power and the distance to the wall follows the free-space path loss equation, rather than the traditional radar equation. Moreover, direct coupling effect is highly mitigated by using highly directional antennas (horn antennas) at the transmitter and receiver of the proposed radar system, which in turn gives a roughly 7m of the maximum detection range. Second, to detect an open doorway with an UGV, a “2-step” doorway detection procedure is utilized in this paper. In the first step (“Step 1”), the nearest wall is detected by using a detection method called Shortest Range Determination (SRD). Then, in the second step (“Step 2”), an open doorway is detected by using the radar and a power-based detection algorithm to scan the space parallel to the nearest wall detected in the first step. This detection algorithm is based on total variation de-noising. It detects an open doorway by detecting the cross-range of the positions of the leftmost and rightmost sides of the open doorway. The experimental results give a detection error of around 20cm. Finally the proposed doorway detection procedure is demonstrated by using the Pioneer 3-AT with the proposed USRP-based radar system. Demonstrations for each step are shown, followed by a complete demonstration of the entire doorway detection procedure. The results show that the detection performance is acceptable.