Novel designs and motion behaviors for small and low-cost mobile robots
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
There is enormous potential for use of small robots in inspection, search-and-rescue, and intelligence–surveillance–reconnaissance (ISR) operations. These (palm sized) robots can navigate through confined, cluttered, and complex environments and thanks to innovative fabrication techniques, they can now be produced at low-cost. If designed appropriately, they can be more readily adopted for real-world applications than traditional robots and they can be deployed in large numbers to increase coverage. Unlike their expensive and complex counterparts, small robots can operate in dangerous or hard to reach spaces where they can be treated as disposable remote sensors. Despite these benefits, several open challenges need to be addressed before miniature robots can achieve their potential and be meaningfully utilized. ☐ Uncertainty brings perhaps the most daunting challenge. Typically, disturbances can be observed by relatively accurate sensors and compensated by feedback. Traditional robots are often assumed to have sufficient control authority to follow trajectories close to their desired reference. However, due to their reduced size and weight, these miniature robots are affected more severely by disturbances: gusts of wind, rough terrain, and even changes in surface friction. With miniaturized actuators, these disturbances cannot be overcome easily, nor can they be measured accurately with low-cost sensors and processors. Environmental disturbances, errors in sensor measurements, and variations between modeled and real-world dynamics are examples of ways that uncertainty affects motion behavior. Because of these effects, combined with the often complex geometry and locomotion mechanisms of small, low-cost robots, their motion is highly unpredictable and cannot be modeled exactly. Considering the challenges of managing these sources of uncertainty, one important question we examine is: how can we principally design models that capture the motion of miniature robots, including its uncertainty, in order to plan control strategies to meet objectives and achieve meaningful tasks? ☐ Some biologically inspired design principles can significantly reduce these effects. This dissertation focuses on novel and reproducible open-access designs to promote real-world use of small, inexpensive robots. The overarching goal is to lower barriers to deployment of miniature robots: through novel designs, models that capture and reproduce their uncertain motion, and principled guidelines for incorporating those new models and motion behaviors in motion planning. To this end, first we design four miniature robots, as openly available platforms that are easily reproduced with reliable motion and manufacturability. By removing the complexity of fabrication, assembly, and maintenance of small mobile robots, this effort can act as a launchpad, expediting new developments in this field. Next, we build a foundation of mathematical tools for describing several experimentally observed motion behaviors in the context of small robots: free-space, wall-following, reflection, and absorption motions. These tools use stochastic differential equations (SDEs) as experimentally validated models with a drift term representative of the robots deterministic (error free) motion and a diffusion term that captures the effect of motion uncertainty on the robot’s position. We develop a strategy for analytically combining heterogeneous primitive motion behaviors and demonstrate, both in simulation and experimentally, how users can predict the probability of success (POS) in motion planning and navigation tasks and improve it by having the robots follow walls or by deploying more robots. Finally, we consider using small robots to intentionally collide with their environment boundaries and show how this behavior can be used in modeling and planning to reduce convergence time.
Description
Keywords
Motion behavior, Low cost, Mobile robot