Fuzzy reactive piloting for continuous driving of long range autonomous planetary micro-rovers

Abstract

Abstract-A complete piloting control subsystem for a highly autonomous long range rover will be defined in order to identify the key control functions needed to achieve continuous driving. This capability can maximize range and number of interesting scientific sites visited during the limited life time of a planetary rover. To achieve continuous driving, a complete set of techniques have been employed: fuzzy based control, real-time artificial intelligence reasoning, fast and robust rover position estimation based on odometry and angular rate sensing, efficient stereo vision elevation maps based on grids, and fast reaction and planning for obstacle detection and obstacle avoidance based on a simple IF-THEN expert system with fuzzy reasoning. To quickly design and implement these techniques, graphical programming has been used to build a fully autonomous piloting system using just the techniques of classic control concepts of cyclic data processing and event driven reaction. Experimental results using the JPL rover Rocky 7 will be given in order to validate the mentioned techniques for continuous driving.

Sun-Synchronous Robotic Exploration: Technical Description and Field Experimentation

“…Some approaches explore descriptions of terrain that encode its variability, e.g. continuous measures of terrain have been implemented as fuzzy sets (Martin-Alvarez et al 1999;Seraji 1999). These representations are generally combined with fuzzy logic controllers which suffer local minima problems similar to potential field approaches.…”

Industrial applications of soft computing: a review

“…They used a complete set of techniques including fuzzy-based control, real-time reasoning, and fast and robust rover position estimation based on odometry, angular rate sensing, and efficient stereo vision [19].…”

Theory and experiments in SmartNav rover navigation

“…Probabilities are precise numbers in the unit interval [0.0, 1.0], and are extracted from statistical analysis of the sensory data. An alternative approach to rover navigation is the use of fuzzy-rule-based methods (see, e.g., Seraji and Howard, 2002;Howard and Seraji, 2001a;Martin-Alvarez et al, 1999;Ojeda et al, 2004;Hagras et al, 2002;Saffiotti, 1997;Tunstel, 1995). Because the fuzzy sets can be linguistic variables, this formulation elevates navigational decisions to a higher level of abstraction and models human-level driving rules expressed in plain English.…”