Compensating for Operational Uncertainty in Man-Machine Systems: A Case Study on Intelligent Vehicle Parking Assist

Abstract

Successful executions of man-machine systems require consistent human operations and reliable machine performances. Compared with the abundant resources on machine reliability improvements, human-related operational uncertainty that has direct impacts on man-machine systems received little attention. Most studies and formal documentations only provide suggestions to alleviate human uncertainty instead of providing specific methods to ensure operation accuracy in real time. In this paper we present a general framework of a reliable system that compensates for human operating uncertainty in realtime. This system learns the response of an operator, constructs the user’s behavior pattern, and then develop new compensated instructions to ensure the completion of the desired tasks, hereby improve the reliability of the entire man-machine systems. The effectiveness of the proposed framework is demonstrated via the development of an intelligent vehicle parking assist. Existing parking assist systems do not account for drivers’ error; neither do these systems consider realistic urban parking spaces with obstacles. Our proposed system computes a theoretical path once a parking space is identified. Audio commands are then sent to the driver with realtime compensation for a minimal deviations from the path. When an operation is too far away from the desired path to be compensated, new set of instructions will be recomputed with the collected uncertainty. Various real-world urban parking scenarios are used to demonstrate the effectiveness of the proposed method. Our system is able to park a vehicle with a space that is as small as 1.07 times the vehicle length with up to 30% uncertainty. Results also show that the compensation scheme not only allows more diverse operators to achieve a desired goal, but also ensures a higher reliability of meeting such goals.