Topic

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Advanced Materials, Human-Machine Interfaces, Trusted AI and Autonomy OBJECTIVE: Successful execution of research and development under this topic will produce a quadruped robot and control architecture that enables faster-than-creep gait in deep muddy environments. DESCRIPTION: Adaptive off-road autonomous locomotion at speed promises disruptive technological overmatch for the future Army. This end state necessitates robust legged locomotion in unforgiving environments, particularly muddy terrain. Authority over legged locomotion in complex environments, particularly quadruped locomotion, has seen rapid acceleration in recent years. Various techniques for navigating uneven, intricate terrain using model predictive control (MPC) and/or reinforcement learning have been presented, energized by the blooming small autonomous robot market and continued interest in planetary exploration [1-4]. Investigators have even tackled the problem of grasses, vines, and entanglements [5]. It is common to describe the terrain that these novel controllers accommodate as “complex”, and authors will often decompose the definition of complex terrain into uneven, dynamic, uncertain, and deformable environments. Yet most of them fall short of claiming effective navigation of truly muddy terrain, and with good reason. Mud and cohesive soils produce widely varying mechanical behaviors under different conditions [6-8]. Attempts to produce legged robotic systems capable of navigating muddy environments typically rely heavily on feet with a large surface area, and using perception to inform a controller about the outcome of a given foot placement [9, 10]. With a clever choice of foot geometry, this design choice largely minimizes the challenge of modeling intrusion, suction, and slipping on even, shallow mud. Future Army applications will require robust control in uneven and deep muddy environments, the ability to overcome suction effects on the gait, and transition to other terrains. Only very recently have research and engineering groups published early efforts to produce a creep gait in a specific “deep mud” environment [11, 12]. As such, the current state of the art for robotic locomotion in mud propels the robot at a speed in the neighborhood of one body length per minute. Many opportunities exist to use these young results as a springboard to a more comprehensive understanding of legged locomotion in cohesive soils, particularly in multidisciplinary spaces involving geophysicists, roboticists, and data scientists. Hybrid (nonlinear MPC and reinforcement learning) and hierarchical control leveraging perceptive and proprioceptive feedback must be brought to bear on this problem, employing new insights into the physics of these challenging terrains and the usefulness of data-driven control. Moreover, limb and robot modification via embodied intelligence informed design could offer insight into the morphologies and passive mechanics necessary to produce adaptability in such systems. Candidate proposals under this topic should, by the end of Phase II, design and implement a novel controller on board a quadrupedal robot capable of traversing muddy environments that produce leg intrusions up to 20% of the leg length. The robot should travel along a path whose distance is at minimum 10 bodylengths at an average speed of 8 bodylengths per minute. PHASE I: This topic is accepting Phase I proposals for a cost up to $250,000 for a 6-month period of performance. Successful applicants should be able to produce a critical mass of preliminary simulation date to demonstrate a thorough and quantitative understanding of structure-soil and fluid-structure-soil interactions associated with gait events in deep mud. Phase I should culminate in a proposed physical robot design for Phase II as well as an in-silico demonstration of a proprioception-only controller rejecting unforeseen perturbations to an 8 body length per minute gait in 20% deep mud. Quantify the efficacy of the novel control architecture against two more traditional baselines. PHASE II: Successful applicants should provide a physical demonstration of a custom autonomous quadruped robot traversing a minimum 10 body length course with up to 20% leg intrusion into deep mud at an average speed of 8 body lengths per minute. Demonstrate the superiority of the novel control and physical design of the robot against one traditional baseline. PHASE III DUAL USE APPLICATIONS: Manufacturing Automation Disaster relief Autonomous support for search and rescue REFERENCES: Griffin et al. “Quadrupedal Walking over Complex Terrain with a Quasi-Direct Drive Actuated Robot.” Special Issue: Robotics Collaborative Technology Alliance (2022) 2:356-384 Liu, Yuqing. Enhancing Quadruped Robot Design with Intelligent Physics-Informed Neural Network-Assisted Dynamic State Estimation and Active Spine Integration. Diss. 2024. Li, Chen, Tingnan Zhang, and Daniel I. Goldman. "A terradynamics of legged locomotion on granular media." Science 339.6126 (2013): 1408-1412. Zhuang, Hongchao, et al. "A Review of Foot–Terrain Interaction Mechanics for Heavy-Duty Legged Robots." Applied Sciences 14.15 (2024): 6541. Yim, Justin K., et al. "Proprioception and reaction for walking among entanglements." 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023. Onyelowe, Kennedy C., et al. "Extensive overview of soil constitutive relations and applications for geotechnical engineering problems." Heliyon 9.3 (2023). Gholizadeh, Esmaeel, and Manoucher Latifi. "A coupled hydro-mechanical constitutive model for unsaturated frictional and cohesive soil." Computers and Geotechnics 98 (2018): 69-81. Zhang, Rui, and Jianqiao Li. "Simulation on mechanical behavior of cohesive soil by Distinct Element Method." Journal of Terramechanics 43.3 (2006): 303-316. Liu, Shipeng, Boyuan Huang, and Feifei Qian. "Adaptation of flipper-mud interactions enables effective terrestrial locomotion on muddy substrates." IEEE Robotics and Automation Letters (2023). Yang, Huaiguang, et al. "Comparative study of terramechanics properties of spherical and cylindrical feet for planetary legged robots on deformable terrain." Journal of Terramechanics 113 (2024): 100968. Godon, Simon, Asko Ristolainen, and Maarja Kruusmaa. "An insight on mud behavior upon stepping." IEEE Robotics and Automation Letters 7.4 (2022): 11039-11046. Godon, Simon, et al. "Walking in Mud: Modelling, Control and Experiments of Quadruped Locomotion." KEYWORDS: Autonomy; Legged Robotics; Human-Machine Integration; Soldier Support; Mechanics; Geophysics; Soil-Structure Interaction; Hybrid Dynamical Systems; Reinforcement Learning; Model Predictive Control; Embodied Intelligence