** 2022 details are left up for visitors to sample the event.**
The Future is Here: Modeling and Simulation Ecosystem of Research and Innovation for Off-Road Mobility and Operations
The Automotive Research Center (ARC) is the U.S. Army’s Center of Excellence for Modeling and Simulation (M&S) of Ground Systems. The Center has gone through several transformations in its history, and most recently there has been a focus on modeling autonomous systems, especially off-road. Through this transformation there has been an accent on partnering with other centers and industry, and leveraging investments. This year, we are in the process of engaging with new partners as well as operationalizing the Ground Vehicle M&S Alliance. In this presentation, we will discuss the vision for the alliance as an ecosystem of modeling and simulation research and innovation. In addition, we will discuss recent developments of capabilities in the ARC as well as share a few success stories of transitions of capabilities.
The Importance of Societal and DEI-based Research in Autonomy
To create autonomic ground vehicles that are capable, reliable, optimal, and survivable in a wide range of changing and challenging environments, we rely on understanding the interactive effects of technological and environmental elements, human factors, and social behavior. Similarly, human and social factors within our teams and in our research process influence what technology we are capable of creating and what impacts that technology can have. Thus, developing the skills to think critically about the human-technology interactions in our research teams and re-imagining how team interactions can be inclusive and supportive are key to moving us closer to the professional and personal goals we hold as engineers.
Fast and Curious: How to Predict and Push Limits of Autonomous Mobility on Deformable Terrains
Making autonomous vehicles able to navigate deformable terrains at operationally-relevant speeds is a critical need for military operations. We envision that achieving such speeds on deformable terrains requires accounting for terrain deformations with high accuracy both in the algorithms that navigate autonomous vehicles, and in the simulations used to develop and evaluate those algorithms. This, however, poses a major research challenge due to the trade-off between high accuracy and computational needs.
This case study is a first step to address this grand challenge with both experimental and simulation components. A novel terrain-aware trajectory planning and control algorithm is deployed and run on an MRZR on soft soil to demonstrate the need and a solution for accurately accounting for deformable terramechanics in planning to safely navigate the vehicle at speed. Data from this demonstration, along with data from soil characterization experiments, is then leveraged to attempt recreation of this experimental vehicle performance in simulation. Two approaches for tire-terrain interaction modeling are utilized with different levels of fidelity, demonstrating the need for high-fidelity terrain representation in evaluating the vehicle navigation and mobility on complex granular terrain and the ARC’s solutions to make it computationally more tractable.
Thus, the results support the vision that it is necessary but also possible to accurately represent deformable terrains in algorithms and simulations to enable high-speed autonomous off-road mobility.
No Time to Fail: Keeping multi-agent off-road teams on the move using a multi-scale hierarchical framework for task allocation and vehicle recharging
Multi-agent off-road team allocation and task completion requires modeling and data analysis across a range of control hierarchies and model fidelities. In this case study, we leverage the outcomes from multiple ARC projects to complete a multi-scale hierarchical framework for task allocation and vehicle recharging. At the low-level, a high-fidelity vehicle and engine model takes vehicle control inputs and simulates the energy consumption, travel velocity, and thermal management of a vehicle and its components. At the mid-level, a single-agent planner considers the goal location and schedule for one agent and generates the travel path based on search-based path planning and model predictive control methods. The high-level team planner calls the single-agent planner and vehicle model to evaluate the path and costs between all task locations and generate a plan for task allocation and coordinated recharging. An ablation study is conducted to evaluate how the task allocation problem changes as a function of the charging behavior, vehicle path plans, and availability of high-fidelity vehicle and engine models. A practical simulation case study is designed to showcase the application of the complete hierarchical system.