Dynamic Modeling and Motion Planning for Robotic Skid-Steered Vehicles - Dynamic Modeling and Motion Planning for Robotic Skid-Steered Vehicles

Skid-steered robots are commonly used in outdoor applications due to their mechanical simplicity, high maneuverability, and robustness. The maneuverability of these robots allows them to perform turning maneuvers ranging from point turns to straight line motion under ideal conditions (e.g., flat terrain and powerful actuators). However, sloped terrain, terrain with high friction, or actuator torque and power limitations can limit the achievable turning radii. The aim of this research is to analyze and experimentally verify the dynamic and power models for skid-steered autonomous ground vehicles equipped with non-ideal (i.e., torque and power limited) actuators and moving on sloped terrains. In particular it investigates the ability of the proposed models to predict motor torques (including motor saturation), power requirement, and minimum turn radius as a function of terrain slope, vehicle heading, payload, terrain parameters and actuator characteristics. The experimental results show that the model is able to predict motor torques for the full range of turning radii on flat ground, i.e., from point turns to straight line motion. In addition, it is shown that the proposed model is able to predict motor torques (including motor saturation) and minimum turn radius as a function of terrain slope, vehicle heading, payload, terrain parameters and actuator characteristics. This makes the model usable for curvilinear motion planning tasks on sloped surfaces. The research uses these results along with Sampling Based Model Predictive Optimization to develop an effective methodology for generating dynamically feasible, energy efficient trajectories for skid-steered autonomous ground vehicles (AGVs) and compares the resultant trajectories with those based on the standard distance optimal trajectories. The simulated and experimental results consider an AGV moving at a constant forward velocity on both wood and asphalt surfaces under various loads. They show that a small increase in the distance of a trajectory over the distance optimal trajectory can result in a dramatic savings in the AGV's energy consumption. They also show that it is not difficult for distance optimal planning to produce trajectories that violate the motor torque constraints for skid-steered AGVs, which can result in poor navigation performance. In addition, the research motivates and provides a methodology that integrates the robot's dynamic model and actuator limitations, and the terrain models with SBMPO to exploit the vehicle momentum as a way to successfully traverse the difficult terrains such as steep hills, mud, or stiff vegetation patches. These scenarios are particularly critical for smaller robots with torque and power limited actuators, which as experimentally shown in this research can easily fail to accomplish their tasks in these environments. In particular, the experimental results showing the efficacy of the proposed methodology are presented for a vegetation patch and a steep hill. Finally, a discussion of the necessary perception work to fully automate the process is included. Further, for walking and running robots, analysis of the power consumption is particularly important for trajectory planning tasks as it enables motion plans that minimize energy consumption and do not violate power limitations of the robot actuators. The research here is motivated by the hypothesis that for certain regimes of operation (i.e., certain gait parameters), legged robots from the RHex family behave in a similar fashion to skid-steered robots while in general curvilinear motion. Hence, using the experience gained from skid-steered wheeled vehicles, presents models of the inner and outer side torques and power requirements for the XRL hexapedal robot. In addition, the applicability of the power model to energy efficient motion planning is illustrated for a walking gait on a vinyl surface.