Low-Cost Dual Radar System for Collision Prediction at Unregulated Intersections

Nair, Dayalan

Low-Cost Dual Radar System for Collision Prediction at Unregulated Intersections

Thesis / Dissertation

2023

Abstract

Automotive collision avoidance has become a key aspect of the automotive design process. The shift in the automotive market towards reducing road accidents has led to the development of many sensor-based collision avoidance systems. The majority of such systems are limited to high-end vehicles. Statistics indicate road intersections are a common location for car accidents. In South Africa, right turns at unregulated intersections require drivers to observe traffic in both directions, which demands substantial attentiveness. This project investigated the suitability of a low-cost bidirectional radar system for determining the safety of turning right at unregulated intersections. The focus was on the detection of four-wheeled vehicles, where frequency-modulated continuous-wave radar facilitated the measurement of the range and speed thereof. An algorithm was designed for detecting multiple targets on a two-lane, bidirectional road using the triangle FMCW waveform. MATLAB simulations ensured selected hardware and processing algorithm met the user requirements. Python allowed for testing real-time processing performance. A roadside test rig consisting of two cameras, two radars, a Raspberry Pi, and a power bank was developed for field testing the proposed system. Performance was validated by comparing radar speed and range estimates to both video footage and GPS estimates. Findings suggested that the final prototype met the project requirements, with a few caveats. The system struggled to detect outer-lane vehicles during peak traffic due to the masking of these vehicles by those travelling on the inner lane. Placing the left radar at an angle reduced the effect of masking and reduced returns from cars travelling away from the host vehicle. The speed estimates had a standard deviation of 8 km/h and 11 km/h for the right side and left side radar, respectively, which was an order of 10 larger than that of GPS estimates.

Keywords

Engineering

Reference:

Collections

Masters

Full item page