This paper describes an implementation of the 3T robot architecture which has been under development for the last eight years. The architecture uses three levels of abstraction and description languages which are compatible between levels. The makeup of the architecture helps to coordinate planful activities with real-time behaviours for dealing with dynamic environments. In recent years, other architectures have been created with similar attributes but two features distinguish the 3T architecture: (1) a variety of useful software tools have been created to help implement this architecture on multiple real robots; and (2) this architecture, or parts of it, have been implemented on a variety of very different robot systems using different processors, operating systems, effectors and sensor suites.

In 1996, NASA will launch the Mars Pathfinder spacecraft, which will carry an 11 kg rover to explore the immediate vicinity of the lander. To assess the capabilities of the rover, the authors have constructed a new microrover testbed consisting of the Rocky 3.2 vehicle and an indoor test arena containing Mars analog terrain and overhead cameras for automatic, real-time tracking of the true rover position and heading. In this paper, the authors present initial performance evaluation results obtained with this testbed. The authors first decompose rover navigation into four major functions: goal designation, rover localization, hazard detection, and path selection. The authors then describe the Mars Pathfinder approach to each function, present results to date of evaluating the performance of each function, and outline their approach to enhancing performance for future missions. The results show key limitations in the quality of rover localization, the speed of hazard detection, and the ability of behavior control algorithms for path selection to negotiate the rock frequencies likely to be encountered on Mars.

We investigate the application of conditional sequencing to robot navigation. Initial experimental results indicate that very robust navigation can be achieved by layering a conditional sequencer on top of a set of simple sensorimotor behaviors. The approach is uniquely flexible, permitting very complex tasks to be programmed reliably in very short periods of time. The technique was used in the recent American Association of Artificial Intelligence mobile robot contest. All of the contest-specific code was written in three days by a single programmer. The robot turned in the best overall performance of any entry. In addition, we present the results of over ninety formal experimental runs performed under a variety of circumstances.>

This paper describes a series of robots developed at JPL to demonstrate the feasibility of using a behavior-control approach to control small robots on planetary surfaces. The round-trip light-time delay makes direct teleoperation of a mobile robot on a planetary surface impossible. Planetary rovers must therefore possess a certain degree of autonomy. However, small robots can only support small computers (due mostly to power, not size constraints). Behavior control provides a means of autonomous control that requires very little computation. The robots described in this paper all used 8-bit, 1-MIP microprocessors with as little as 4 k and no more than 40 k of memory, and extremely simple sensors. Despite these limitations they reliably perform both autonomous navigation and manipulation in both indoor and outdoor rough-terrain environments.>

The author introduces ALFA, a language for programming reactive control mechanisms for autonomous mobile robots. ALFA programs consist of networks of computational modules connected by communications channels. These channels are themselves computational entities which perform mediation tasks. This makes it possible to design modular behaviors which may be inserted and removed without having to restructure the interconnections among modules and allows the system to interface cleanly with higher-level reasoning processes. ALFA supports a dataflow computation model as well as a state-machine model. The results of two experiments using ALFA to control real robots are briefly discussed.>

Reliable goal-directed navigation is demonstrated on a real robot in an unmodified office environment using crude sensors and minimal computation. The control system is based on reactive behaviors which fail cognizantly, that is, which can detect when something goes wrong. Cognizant failure allows unreliable behaviors to be combined by a reactive scheduler to produce an overall behavior which is more reliable than any of its components. Algorithms which fail cognizantly are much easier to produce than algorithms which never fail. Several such algorithms are presented.>

A path planner and an execution monitoring planner that will enable the rover to navigate to its various destinations safely and correctly while detecting and avoiding hazards are described. An overview of the complete architecture is given. Implementation and testbeds are described. The robot can detect unforeseen obstacles and take appropriate action. This includes having the rover back away from the hazard and mark the area as untraversable in the rover's internal map. The experiments have consisted of paths roughly 20 m in length. The architecture works with a large variety of rover configurations with different kinematic constraints.>