This paper proposes an improved version of 3D pure proportional navigation (PPN) against a manoeuvring target. The main research hypothesis is that the performance of 3D PPN can be improved by properly selecting the direction of the guidance command as there exists an infinite number of potential directions complying with the PPN concept in 3D space. Analysis on the relative motion confirms the validity of the hypothesis and leads to the development of a new guidance algorithm. Unlike traditional 3D PPN, the guidance algorithm developed adapts the direction, but maintains the magnitude of the commanded acceleration proportional to only the line-of-sight (LOS) rate. The validity and performance of the proposed guidance algorithm are investigated through theoretical analysis and numerical simulations. Index Terms 3D PPN, manoeuvring target, direction of commanded acceleration, relative motion analysis I. INTRODUCTION Pure proportional navigation (PPN) guidance law [1]-[6] is a major class of proportional navigation (PN) guidancelaws and mainly used for endo-atmospheric interception, whereas true proportional navigation (TPN) guidance law [7]-[9] is another major PN class that is commonly used for exo-atmospheric interception. The commanded acceleration vector of PPN is perpendicular to the interceptor's velocity vector and its magnitude is proportional to the line-of-sight (LOS) angular rate. PPN is preferred over many other guidance laws mostly thanks to its robustness and practicality [10]. Implementation of PPN mainly requires the measurement of LOS rate, which is generally available from the gimballed seeker system on the interceptor.It is known that PPN provides excellent capturability against non-manoeuvring target for endo-atmospheric interception [6]. The LOS rate and commanded acceleration of PPN are continuously decreasing during the guidance process, and the capture region is extremely large. However, if the target is manoeuvring with large acceleration, performance of PPN might be significantly degraded. Many researchers have investigated the performance of PPN against manoeuvring targets using linear or nonlinear methods. For example, Shukla and Mahapatra [11] extended