Evaluation of gaming environments for mixed reality interfaces and human supervisory control in telerobotics

Abstract

Telerobotics refers to a branch of technology that deals with controlling a robot from a distance. It is commonly used to access difficult environments, reduce operating costs, and to improve comfort and safety. However, difficulties have emerged in telerobotics development. Effective telerobotics requires maximising operator performance and previous research has identified issues which reduce operator performance, such as operator attention being divided across the numerous custom built interfaces and continuous operator involvement in a high workload situation potentially causing exhaustion and subsequent operator error. This thesis evaluates mixed reality and human supervisory control concepts in a gaming engine environment for telerobotics. This concept is proposed in order to improve the effectiveness of current technology in telerobotic interfaces. Four experiments are reported in this thesis which covers virtual gaming environments, mixed reality interfaces, and human supervisory control and aims to advance telerobotics technology. This thesis argues that gaming environments are useful for building telerobotic interfaces and examines the properties required for telerobotics. A useful feature provided by gaming environments is that of overlying video on virtual objects to support mixed reality interfaces. Experiments in this thesis show that mixed reality interfaces provide useful information without distracting the operator from the task. This thesis introduces two response models based on the planning process of human supervisory control: Adaptation and Queue response models. The experimental results show superior user performance under these two response models compared to direct/manual control. In the final experiment a large number of novice users, with a diversity of backgrounds, used a robot arm to push blocks into a hole by using these two response models. Further analyses on evaluating the user performance on the interfaces with two response models were found to be well fitted by a Weibull distribution. Operators preferred the interface with the Queue response model over the interface with the Adaptation response model, and human supervisory control over direct/manual control. It is expected that the increased sophistication of control commands in a production system will usually be greater than those that were tested in this thesis, where limited time was available for automation development. Where that is the case the increases in human productivity using human supervisory control found in this experiment can be expected to be greater. The research conducted here has shown that mixed reality in gaming environments, when combined with human supervisory control, offers a good route for overcoming limitations in current telerobotics technology. Practical applications would benefit by the application of these methods, making it possible for the operator to have the necessary information available in a convenient and non-distracting form, considerably improving productivity.