A Case Study in System-Level Physics-Based Simulation of a Biomimetic Robot

作者：Primerano Richard^*; Wilkie David; Regli William C

来源：IEEE Transactions on Automation Science and Engineering, 2011, 8(3): 664-671.

DOI：10.1109/TASE.2011.2106774

• Biorobotics
• legged locomotion
• modeling methodologies
• simulation

Biomimetic robots are a new and challenging frontier for robotic systems. Designs inspired by nature are creating new approaches to problems such as to planetary surface exploration, minimally invasive surgery, or inspection of piping and cabling. However, these new biomimetic robotic systems are a challenge to design, simulate, and control.
This paper presents a case-study in the design and physics-based simulation of a unique snake-inspired robot. The Drexel Snake Robot is a novel hybrid capable of both undulatory and rectilinear motion. In order to design gaits, test control algorithms and perform path planning for this robot, we develop a system-level physics-based simulation that captures engineering phenomena across many disciplines: mechanical, electrical, software, electronics, and the robot's external environment.
As the snake robot moves, there is considerable slippage between its feet and the ground. Consequently, contact and friction forces play a significant role in dictating the path followed by the robot for a given set of joint motions. A closed-form equation (or set of equations) describing the robot's motion in this environment cannot be derived in a simple manner. Consequently, the material presented here are based the comparison of experimental and simulation results.
While developing the simulation model, we detail the process of extracting necessary physical, kinematic and dynamics properties directly from the robot's computer-aided design. This process is not straightforward and the paper documents the issues and lessons learned that will be of use to others wishing to create full virtual models for their systems. Finally, we show how to calibrate the simulation model for fidelity and accuracy with several examples showing that it can be used to test gait and mobility patterns and identify nonobvious secondary phenomena to emerge from the design.
Note to Practitioners-Anyone wishing to develop a physically-based simulation of a robot, especially a biorobotic system, directly from CAD needs to address the issues we present in this paper. Our belief is that the process we describe is generally applicable and that the issues we address (in particular, in how the CAD data is extracted and integrated into simulation) is a source for many new problems in systems integration and standards.

• 出版日期2011-7