Model-Based Reliable Mission Planning of Off-Road Autonomous Ground Vehicles Under Uncertain Environments
Yixuan Liu · Deep Blue (University of Michigan) · 2022
Off-road autonomous ground vehicles (AGVs) and other types of robotics are drawing increased attention in recent years as they are able to replace human in dangerous or boring working environments such as in the presence of wildfires or earthquakes, battlefield, and agricultural field. Mission planning of off-road AGVs plays a vital role in ensuring the successful and autonomous operation of AGV in the off-road environment. The unstructured off-road environment and various sources of uncertainty pose challenge to mission planning of off-road AGVs. This research focuses on mission planning of off-road AGVs with mobility reliability considerations. The goal is to identify a path that is not only shortest in terms of travel distance, but also reliable to ensure the success of a mission. It is a systematic work that contains three main steps: (i) vehicle mobility modeling, (ii) mobility reliability analysis, and (iii) reliability-based path planning. This dissertation aims to provide novel approaches to properly perform the three steps. We firstly construct a dynamic ensemble of Nonlinear Autoregressive Network with Exogenous inputs (NARX) models over time to accurately prediction mobility of off-road AGVs. Secondly, a simulation-based mission mobility reliability (MMR) analysis framework is developed to account for uncertainty in mobility prediction of off-road AGVs in mission planning phase and a dynamic updating scheme is proposed to update the MMR estimation using online mobility data. Then the adaptive surrogate modeling is used to calculate the state mobility reliability and incorporate the soil information into mobility reliability map. A reliability-based path planning method is developed for both single and multiple vehicle using a Physarum-based algorithm and the navigation map. Finally, a rapidly-exploring random trees star (RRT*) algorithm is studied to account for multiple the mission mobility reliability (MMR) constraints in mission planning path. Multiple case studies are used to demonstrate the effectiveness of the proposed approaches. The results show that the proposed approaches can effectively identify optimal paths while satisfying certain mission mobility reliability requirements.