Providing situation awareness in complex multi-vehicle operations

Abstract

This thesis addresses the problem of safety in complex multi-vehicle interactions, initially in the domain of the mining industry. Operating a vehicle in a mining environment can be particularly dangerous due to the difficult environment, lack of visibility from the vehicle cabin, long shifts and many other factors that result in many accidents and near misses each year. The state of the art in vehicle safety systems do not address all of these problems. Current systems lack the ability to effectively determine the risks and threats that are important, and to be able to communicate these to an operator in a complex situation. This thesis introduces the concept of situation awareness into the mining domain. Situation awareness provides a model of the way humans take in information from their surroundings, process this information and predict the future state of the environment. Human error in vehicle accidents can be attributed to deficiencies in one or more of these components. The aim of a safety system is to improve situation awareness, while minimising or eliminating the negative impact on situation awareness caused by the introduction of the technology. The requirements for determining and distributing vehicle state information are examined and tested. The concept of context is introduced as a means of filtering out unnecessary information and false alarms to the operator, as well as providing higher level information that requires less mental bandwidth to process. This is important as false alarms and unnecessary information annoys the operators and distracts them from the important task of controlling the vehicle. The quality of the vehicle state information is shown to be improved using localisation algorithms introduced in this thesis. These algorithms utilise the context information, digitised maps and a model of the vehicle dynamics to estimate and predict vehicle state information. The theory and testing of these algorithms are an important contribution of this thesis. The methods of providing feedback to the operator are introduced, and the role of assistive, reactive and preventative feedback is discussed. A measure of vehicle safety is provided using metrics based on the known safe rules of operation within a mine. An important contribution of this thesis is the real time feedback of safety performance based on these metrics which are known to contribute to the risk of an accident. Comprehensive experimental results are provided for the concepts introduced in this thesis from system installations in several mines in Australia and Indonesia.