Terrain Aided Localisation of Autonomous Vehicles in Unstructured Environments

Abstract

This thesis is concerned with the theoretical and practical development of reliable and robust localisation algorithms for autonomous land vehicles operating at high speeds in unstructured, expansive and harsh environments. Localisation is the ability of a vehicle to determine its position and orientation within an operating environment. The need for such a localisation system is motivated by the requirement of developing autonomous vehicles in applications such as mining, agriculture and cargo handling. The main drivers in these applications are for safety, efficiency and productivity. The approach taken to the localisation problem in this thesis guarantees that the safety and reliability requirements imposed by such applications are achieved. The approach also aims to minimise the engineering or modification of the environment, such as adding artificial landmarks or other infrastructure. This is a key driver in the practical implementation of a localisation algorithm.

In pursuit of these objectives, this thesis makes the following principal contributions:

1. The development of an Iterative Closest Point - Extended Kalman Filter (ICP-EKF) algorithm - a map-based iconic algorithm that utilises measurements from a scanning laser rangefinder to achieve localisation. The ICP-EKF algorithm entails the development of a map-building algorithm. The main attraction of the map-based localisation algorithm is that it works directly on sensed data and thus does not require extraction and matching of features. It also explicitly takes into account the uncertainty associated with measurements and has the ability to include measurements from a variety of different sensors.

2. The development and implementation of an entropy-based metric to evaluate the information content of measurements. This metric facilitates the augmentation of landmarks to the ICP-EKF algorithm thus guaranteeing reliable and robust localisation.

3. The development and adaptation of a view-invariant Curvature Scale Space (CSS) landmark extraction algorithm. The algorithm is sufficiently robust to sensor noise and is capable of reliably detecting and extracting landmarks that are naturally present in the environment from laser rangefinder scans.

4. The integration of the information metric and the CSS and ICP-EKF algorithms to arrive at a unified localisation framework that uses measurements from both artificial and natural landmarks, combined with dead-reckoning sensors, to deliver reliable vehicle position estimates.

The localisation framework developed is sufficiently generic to be used on a variety of other autonomous land vehicle systems. This is demonstrated by its implementation using field data collected from three different trials on three different vehicles. The first trial was carried out on a four-wheel drive vehicle in an underground mine tunnel. The second trial was conducted on a Load-Haul-Dump (LHD) truck in a test tunnel constructed to emulate an underground mine. The estimates of the proposed localisation algorithms are compared to the ground truth provided by an artificial landmark-based localisation algorithm that uses bearing measurements from a laser. To demonstrate the feasibility and reliability of both the natural landmark extraction and localisation algorithms, these are also implemented on a utility vehicle in an outdoor area within the University's campus. The results demonstrate the robustness of the proposed localisation algorithms in producing reliable and accurate position estimates for autonomous vehicles operating in a variety of unstructured domains.

About the Author

Rajmohan Madhavan was born in Madras, India, in 1973. He received the Bachelor of Engineering degree (Electrical and Electronics) in 1995 from the College of Engineering, Guindy, Anna University, India and the Master of Engineering (Research) degree (Control and Robotics) in 1997 from the Department of Engineering, The Australian National University, Australia. He received his Ph.D. degree (Field Robotics) in 2001 from the School of Aerospace, Mechanical and Mechatronic Engineering (Department of Mechanical and Mechatronic Engineering at the time of completion) from the University of Sydney, Sydney, Australia. Since 2001, he has been a research associate with the Oak Ridge National Laboratory (ORNL) and is currently a guest researcher with the Intelligent Systems Division of the National Institute of Standards and Technology (NIST). His current research interests include autonomous vehicle navigation in large and complex environments, distributed heterogenous sensing and systems and control theory.