SquadBot v2: High Performance Humanoid for Urban Exploration
Abstract
Summary of Innovation: We propose to research, develop, and demonstrate SquadBot v2, a humanoid robot for operating in urban environ,ments alongside soldiers as a robotic squad member. Humanoid robots like SquadBot promise to enable revolutionary changes in urban t,actics across the full spectrum of squad operations, including neighborhood watch, building search, patrol, street combat, and build,ing clearing. While humanoid robot mobility and utility have made great strides, they are still subpar compared to real humans, bein,g much slower and requiring extensive teleoperation. To address these deficits, we will improve our behavior architecture to include, persistence, particularly for moving obstacles that the robot may encounter during exploration. We will also develop advanced mobil,ity algorithms focused on multi-contact locomotion, including approaches for bracing, crawling, and standing up again after a fall.,SquadBot v2 will be a high performance, next generation humanoid robot, with physical capabilities approaching that of a human. The,robot will be capable of multi-contact locomotion and will be able to remove debris and obstacles that block its path. SquadBot v2 w,ill be able to survive some falls and stand up from the ground. It will have one of the highest power-to weight ratios of any humano,id robot in existence, enabling fast, dynamic motions. SquadBot v2 will be operated through various means, including remote user int,erfaces. We will focus on improving our variable autonomy framework for both fully autonomous behaviors and human-in-the-loop behavi,ors, with an eye towards how each modality contributes to maximizing speed and reliability. We will maximize success by allowing ope,o-actively analyze limitations to performance throughout the system and determine improvements that could impact results.The success, of this project will be enabled by the following key innovations: SquadBot v2 robot design to support arbitrary environmental conta,ct. We will investigate novel joint, structure, and actuator designs for SquadBot v2. Our goal will be to maintain the high range of, motion of SquadBot v1, while better enabling fall survivability and multi-contact. We will achieve this with onboard power, hydraul,ic power, computer, and batteries. We will focus on using composite shells to provide survivability, as well as surfaces that can co,ntact the environment. Multi-contact whole body control for building exploration. Multi-contact locomotion is extremely challenging,due partly to a lack of simplified models for planning and control of multi-contact motions. In this project, we will extend our wal,king and balance algorithms to include the knees, hands, and forearms for balance. We will develop techniques for perceiving availab,le foot and handholds, estimating contact conditions, determining motion feasibility, and planning motion across degraded indoor ter,rains terrain. Persistent behaviors for building exploration. We will improve our fast automated robot behaviors, which will include,@##@00025@#,earch, identify humans, and go up and down stairs. We will explore techniques to combine these skill-level behaviors into task-level,performance. Autonomous debris clearing for building exploration. We will work with the University of Washington to integrate and im,prove their autonomous picking and clearing manipulation framework. We will focus on extending these existing algorithms to legged,,bimanual systems. We will also seek to incorporate the ability for user interaction in the planning process, such that the operator,can intervene and assist at any point.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 13, 2022
- Source ID
- N000142212593
Entities
People
- Robert G. Griffin
Organizations
- Florida Institute for Human and Machine Cognition
- Office of Naval Research
- United States Navy