Measuring worst-case errors in a robot workcell
- Sandia National Labs., Albuquerque, NM (United States). Intelligent Systems and Robotics Center
Errors in model parameters, sensing, and control are inevitably present in real robot systems. These errors must be considered in order to automatically plan robust solutions to many manipulation tasks. Lozano-Perez, Mason, and Taylor proposed a formal method for synthesizing robust actions in the presence of uncertainty; this method has been extended by several subsequent researchers. All of these results presume the existence of worst-case error bounds that describe the maximum possible deviation between the robot`s model of the world and reality. This paper examines the problem of measuring these error bounds for a real robot workcell. These measurements are difficult, because of the desire to completely contain all possible deviations while avoiding bounds that are overly conservative. The authors present a detailed description of a series of experiments that characterize and quantify the possible errors in visual sensing and motion control for a robot workcell equipped with standard industrial robot hardware. In addition to providing a means for measuring these specific errors, these experiments shed light on the general problem of measuring worst-case errors.
- Research Organization:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 541822
- Report Number(s):
- SAND-97-2340; ON: DE98000892; BR: DP0102021; CRN: C/SNL--SC9201118; TRN: 98:008524
- Resource Relation:
- Other Information: PBD: Oct 1997
- Country of Publication:
- United States
- Language:
- English
Similar Records
Artificial awareness for robots using artificial neural nets to monitor robotic workcells
Rethinking synthesis of fine-motion strategies