Tracking Filter Design Methodologies and Radar Applications  

Background

In radar sensing, it is important to use not only measurements at a single time instant but also temporal information. The position and velocity of a target change continuously, and a tracking filter can combine past measurements with a motion model to estimate the target state while reducing the influence of measurement noise.

Typical state-estimation methods include Kalman filters, extended Kalman filters, unscented Kalman filters, and particle filters. When multiple targets are observed simultaneously, additional problems arise, such as data association, multi-target tracking, clutter rejection, false alarms, and missed detections.

Radar has a unique advantage in that it can directly observe Doppler velocity in addition to range and angle. Therefore, an important research topic is how to effectively incorporate Doppler velocity information into tracking-filter design.

Research Objectives

This research aims to develop tracking filters and state-estimation methods that exploit the characteristics of radar measurements. In recent years, tracking filters have increasingly been constructed empirically using machine learning and data-driven approaches. Our work also aims to clarify the theoretical properties of such filters and to establish design methodologies with analytical support.

Main topics include:

• Moving-object tracking using Kalman filters

• State-space modeling suitable for radar measurements

• Position and velocity estimation using range, angle, and Doppler velocity

• Multi-target tracking

• Robust tracking under outliers, clutter, and missed detections

• Sensor fusion using radar and other sensors

• Applications to human tracking, pedestrian detection, and ego-motion estimation

• Performance analysis and parameter design of tracking filters

Related Work

Our previous work has addressed Kalman-filter design for radar measurements, performance analysis of tracking filters, moving-object tracking using Doppler velocity, multi-target tracking, and applications to ego-motion estimation.

In particular, the research focuses on how measurement noise and target motion models affect tracking performance. It also investigates design guidelines and performance evaluation methods for tracking filters. By incorporating velocity information obtained from radar, the goal is to achieve more stable state estimation than tracking methods based only on position measurements.

These studies are closely related to short-range radar-based human sensing, pedestrian tracking around vehicles, self-localization using millimeter-wave MIMO radar, and environment perception using multiple sensors.

Selected Publications

• Ryuto Terawake, Keiji Jimi, and Kenshi Saho, "Millimeter-wave Radar-based Ego-Trajectory Estimation of a Vehicle Using Multi-Target Tracking Filter," IEEE Sensors Letters, vol. 10, art no.3501804 , March 2026. DOI: 10.1109/LSENS.2026.3669681

• Kenshi Saho and Takanori Shibata, "Closed-form steady-state bias error of tracking filter using coordinated turn model for constant velocity target," IEICE Communications Express, vol. 11, no. 5, pp. 245-250, May 2022. 

• Manami Seta and Kenshi Saho, "Best-acceleration filter for moving object tracking," CEUR Workshop Proceedings, vol. 4082, pp.29-38, September 2025. DOI: 10.5281/zenodo.19583088

• Kenshi Saho, "Kalman Filter for Moving Object Tracking: Performance Analysis and Filter Design," Book Chapter in Kalman Filters - Theory for Advanced Applications, Chapter 12, pp.233-251, ISBN 978-953-51-5618-5, Ginalber Luiz de Oliveira Serra (Ed.), InTechOpen, February 2018.

• Kenshi Saho and Masao Masugi, "Automatic Parameter Setting Method for an Accurate Kalman Filter Tracker Using an Analytical Steady-State Performance Index," IEEE Access, vol.3, pp.1919–1930, October 2015.

Keywords

tracking filter, Kalman filter, state estimation, moving-object tracking, multi-target tracking, data association, radar tracking, Doppler velocity, sensor fusion, pedestrian tracking, human tracking, ego-motion estimation, self-localization, radar signal processing