A low-cost indoor testbed for multirobot adaptive navigation research

摘要

Conventional navigation typically focuses on guiding a vehicle along a pre-defined path or through a series of waypoints. In contrast, Adaptive Navigation techniques determine how to move based on environmental measurements collected during the navigation process. The most studied version of this is navigation through a scalar field, where a single point measurement such as temperature is associated with each point in the field, in order to find the location of minimum or maximum value. Many proposed adaptive navigation techniques use multiple robots, exploiting their spatial distribution in order to improve the navigation process. The potential of such methods is compelling, enabling navigation to/along a variety of critical field features relevant to a wide range of applications, and doing so more efficiently than conventional navigation techniques. Unfortunately, most of these methods have not advanced beyond the simulation phase given the challenge of having access to a convenient experimental testbed that can support maturation of these techniques and their experimental verification. In addition to supporting multirobot operation, such a testbed should allow the scalar field of interest to be explicitly specified and customized in order to evaluate the navigation task of interest. To address this challenge, our team has developed a low-cost indoor multirobot adaptive navigation testbed with several intriguing features; these include a suite of 12 small omnidrive robots tracked via an optical tracking system, on-board sensors that measure the reflectivity of a printed mat on which the robots drive and which graphically represents a scalar field of interest, and a virtual reality visualization system for graphically presenting complex scalar fields and navigation performance. This paper reviews adaptive navigation, summarizes challenges in establishing experimental testbeds, and presents the design and functionality of our testbed.