Performance Assessment of an Integrated Radar Architecture for Multi-Types Frontal Object Detection for Autonomous Vehicle Fakhrul Razi Ahmad Zakuan * , Umar Zakir Abdul Hamid * , Dilip Kumar Limbu * , Hairi Zamzuri † , and Muhammad Aizzat Zakaria ‡ * Moovita Pte Ltd, 8 Burn Road Trivex Building, 13-01, 369977 Singapore † Vehicle System Engineering iKohza, Malaysia-Japan International Institute of Technology, Universiti Teknologi Malaysia, 54100, Kuala Lumpur Malaysia ‡ Faculty of Manufacturing Engineering, Universiti Malaysia Pahang, 26000 Pekan, Pahang, Malaysia. Email: razi@moovita.com, umartozakir@gmail.com Abstract—A hazard-free autonomous vehicle (AV) navigation requires the presence of a dependable perception module. For the frontal object and environment detection, the perception module needs to be able to provide a comprehensive detection region to allow blind spot monitoring. In this work, an integrated perception architecture for frontal object detection of an AV is introduced. The architecture is an integration of central, left and right sides radar for the frontal area. The radar placement of the design is expected to reduce the frontal blind spot region of the vehicle. The architecture’s performance is analyzed by assessing each of the radar’s performance. This includes their respective ability to measure multi-type of obstacles (pedestrian, black car and motorcycle) as well as evaluating the max range detection of the radars. The results show that each of the radar managed to detect multi-type obstacles as well as possessing a maximum detection range, which will allow for a reduced blind spot area and aids the hazard mitigation actions. The findings of this preliminary study are important to develop a more comprehensive perception module of an autonomous vehicle. I. I NTRODUCTION The arrival of the Fourth Industrial Revolution saw the emergence of various types of new technologies with dis- ruptive effects. These technologies stimulate the formation of new markets and subsequently affect the current support network to the available technology. Among the examples are driverless vehicle [1]. Major carmakers are racing towards the implementation of a fully autonomous vehicle (AV). Spin-off companies are founded by these giant companies. For exam- ples, Volvo is collaborating with Autoliv by setting up Zenu- ity, and Audi is establishing Autonomous Intelligent Driving, which solely focuses on the automated driving field research and development [2] [3]. A well-built AV navigates by compensating the absence of the driver in its feature with the incorporation of several submodules, including the perception module, which monitors the host vehicle environment. Light Detection and Ranging (LiDAR) and Radio Detection and Ranging (Radar) are the most frequently utilized perception devices for the AV development [4]. However, compared to Radar, LiDAR possessed several disadvantages with regards to its ability to perform in varied types of weather, particularly heavy raining [5] [6]. To allow the AV to be used in the tropical countries, such as Malaysia and Singapore, where the annual rainfall intensity level is high, radar incorporation into the perception module would be helpful. Furthermore, due to the dynamic nature of the road hazards which is occupied by many types of road users and objects, radar as part of the perception modules should be able to detect multi-type road objects and hindrances. In addition, for the navigation in crowded areas, where previously occluded pedestrians could suddenly appear from the blind spot region, radar should be able to detect pedestrians too [7]. Thus, to allow for a reliable detection feature of the AV and compensate the absence of the human driver, the development of the AV perception modules should consider all of these factors. It is important to be mentioned that the work will only focus on the frontal detection region of an AV, where future works will encompass of the rear and side detection regions discussion. A. Objective of This Work Based on the brief literature in the previous section, in this work, a comprehensive perception architecture utilizing radar is proposed. The design, incorporated of three radars, is constructed to enable a reduced amount of blind spot area for the frontal object detection. The radar architecture perfor- mance is evaluated and measured. The evaluation analysis encompassed of the ability of the vehicle to measure the varied type of object, as well as their respective max range detection distance. Since this work is done in Malaysia and Singapore, the ability of the radar is measured in its ability to detect cars, motorcycles and pedestrians, the three most frequent road users in Southeast Asia (ASEAN) region [8]. As motorcycle is not heavily utilized in Europe, thus this study is important to address the said topics in relation to ASEAN requirement. Due to the brief nature of this paper, the main objective of this paper is to evaluate and report each radar’s performance of the design in measuring the environment. From the findings, the weakness is analyzed