Page 59 - Autonomous Mobile Robots

42 Autonomous Mobile Robots

2.6.2 Merits of the Proposed Algorithm over Other Feature
Extraction Techniques ........................................ 74
2.7 Multiple Line-of-Sight Targets — RADAR Penetration............. 76
2.8 RADAR-Based Augmented State Vector............................. 79
2.8.1 Process Model ................................................. 80
2.8.2 Observation (Measurement) Model .......................... 84
2.8.2.1 Predicted power observation formulation ......... 85
2.9 Multi-Target Range Bin Prediction — Results....................... 89
2.10 Conclusions ............................................................ 93
Acknowledgments ............................................................ 94
References .................................................................... 95
Biographies ................................................................... 97

2.1 INTRODUCTION
Current research in autonomous robot navigation [1,2] focuses on mining,
planetary-exploration, fire emergencies, battlefield operations, as well as
on agricultural applications. Millimeter wave (MMW) RADAR provides
consistent and accurate range measurements for the environmental ima-
ging required to navigate in dusty, foggy, and poorly illuminated envir-
onments [3]. MMW RADAR signals can provide information of certain
distributed targets that appear in a single line-of-sight observation. This
work is conducted with a 77-GHz frequency modulated continuous wave
(FMCW) RADAR which operates in the MMW region of the electromagnetic
spectrum [4,5].
For localization and map building, it is necessary to predict the target loc-
ations accurately given a prediction of the vehicle/RADAR location [6,7].
Therefore, the first contribution of this chapter offers a method for pre-
dicting the power–range spectra (or range bins) using the RADAR range
equation and knowledge of the noise distributions in the RADAR. The
predicted range bins are to be used ultimately as predicted observations
within a mobile robot RADAR-based navigation formulation. The actual
observations take the form of received power/range readings from the
RADAR.
The second contribution of this chapter is an algorithm which makes optimal
estimates of the range to multiple targets down-range, for each range spectra
based on received signal-to-noise power. We refer to this as feature detec-
tion based on target presence probability. Results are shown which compare
probability-based feature detection with other feature extraction techniques
such as constant threshold [9] on raw data and constant false alarm rate (CFAR)

© 2006 by Taylor & Francis Group, LLC

FRANKL: “dk6033_c002” — 2006/3/31 — 17:29 — page 42 — #2