Erann Gat

Remote agent prototype for spacecraft autonomy

NASA has recently announced the New Millennium Program (NMP) to develop faster, better, cheaper spacecraft in order to establish a virtual presence in space. A crucial element in achieving this vision is onboard spacecraft autonomy, requiring us to automate functions which have traditionally been achieved on ground by humans. These include planning activities, sequencing spacecraft actions, tracking spacecraft state, ensuring correct functioning, recovering in cases of failure and reconfiguring hardware. In response to these challenging requirements, we analyzed the spacecraft domain to determine its unique properties and developed an architecture which provided the required functionality. This architecture integrates traditional real-time monitoring and control with constraint-based planning and scheduling, robust multi-threaded execution, and model-based diagnosis and reconfiguration. In a five month effort we successfully demonstrated this implemented architecture in the context of an autonomous insertion of a simulated spacecraft into orbit around Saturn, trading off science and engineering goals, and achieving the mission goals in the face of any single point of hardware failure. This scenario turned out to be among the most complex handled by each of the component technologies. As a result of this success, the integrated architecture has been selected to control the first NMP flight, Deep Space One, in 1998. It will be the first AI system to autonomously control an actual spacecraft.

Towards principled experimental study of autonomous mobile robots

We review the current state of research in autonomous mobile robots and conclude that there is an inadequate basis for predicting the reliability and behavior of robots operating in unengineered environments. We present a new approach to the study of autonomous mobile robot performance based on formal statistical analysis of independently reproducible experiments conducted on real robots. Simulators serve as models rather than experimental surrogates. We demonstrate three new results: 1) Two commonly used performance metrics (time and distance) are not as well correlated as is often tacitly assumed. 2) The probability distributions of these performance metrics are exponential rather than normal, and 3) a modular, object-oriented simulation accurately predicts the behavior of the real robot in a statistically significant manner.

Exploiting known topologies to navigate with low-computation sensing

Exploiting prior knowledge about the general characteristics of an environment can reduce the amount of sensing required. For example, in an indoor environment floors tendnto be flat and walls tend to be straight and static. In such an environment, simple range sensors can provide enough information to support robust sensor-driven goal-directed navigation. This paper will describe a navigation experiment using a real robot, and speculate on how the techniques used can be extended to other domains.

Autonomy software verification and validation might not be as hard as it seems

The verification and validation of autonomy software is widely believed to be a challenging unsolved problem. To a certain extent this is true, but in this paper I argue that the problem is not nearly as severe as seems to be widely perceived. Many of the perceived hard problems in autonomy software V&V also exist for traditional software, and can be solved using many of the same methods and techniques used for traditional spacecraft software. In particular, the problem of intractably large state spaces exists for any nontrivial software system. This problem can be addressed for autonomy software in the same way that it has been addressed for traditional software: by decomposing the large state space into a tractable number of equivalence classes that exhibit qualitatively identical behavior, each one containing a large number of states.

Simple sensors for performing useful tasks autonomously in complex outdoor terrain

This paper describes the control system for Rocky IV, a prototype microrover designed to demonstrate proof-of-concept for a low-cost scientific mission to Mars. Rocky IV uses a behavior-based control architecture which implements a large variety of functions displaying various degrees of autonomy, from completely autonomous long-duration conditional sequences of actions to very precisely described actions resembling classical AI operators. The control system integrates information from infrared proximity sensors, proprioceptive encoders which report on the state of the articulation of the rovers suspension system and other mechanics, a homing beacon, a magnetic compass, and contact sensors. In addition, significant functionality is implemented as virtual sensors, computed values which are presented to the system as if they were sensors values. The robot is able to perform a variety of useful tasks, including soil sample collection, removal of surface weathering layers from rocks, spectral imaging, instrument deployment, and sample return, under realistic mission- like conditions in Mars-like terrain.

Continuous motion using task-directed stereo vision

The performance of autonomous mobile robots performing complex navigation tasks can be dramatically improved by directly expensive sensing and planning in service of the task. The task-direction algorithms can be quite simple. In this paper we describe a simple task-directed vision system which has been implemented on a real outdoor robot which navigates using stereo vision. While the performance of this particular robot was improved by task-directed vision, the performance of task-directed vision in general is influenced in complex ways by many factors. We briefly discuss some of these, and present some initial simulated results.

Wavelength division multiplexed fiber optic absolute position encoder

A method of wavelength division multiplexing (WDM) for fiber optic sensors using a broadband light source and narrow bandpass thin film optical filter coatings on cylindrical graded index lenses has been developed. The WDM system is characterized by parallel information from all channels in real time. The system is being applied to a digital, rotary, absolute position encoder.