Projects

Bioinspiration

The focus of this work, in partnership with the Air Force Center of Excellence on Nature-Inspired Flight Technologies and Ideas (NIFTI) at UW, is to build a process for translating biological capabilities for agile flight in dynamic, complex and unknown environments to appropriate designs and algorithms for engineered flight vehicles. In engineered systems, sensors are typically complex in terms of computational requirements, weight and physical design and have typically been designed to provide data on individual quantities with high density. Conversely, biological systems employ a high number of simple sensors that provide data for limited portions of a quantity of interest and which must be fused across both spatial and time scales.

These biological systems demonstrate the ability to fly effectively in highly cluttered environments such as under the forest canopy, safely land on variable and moving terrain (e.g. branches or vertical walls), operate with highly variable lighting and acoustic conditions, and achieve desired behaviors in the presence of extreme environmental perturbations such as wind and adversaries.

Morphing vehicles

Autonomous agile flight of aircraft and spacecraft with dynamic payloads and interacting with the environment is a technological goal of high immediate interest in both commercial and military applications. We are exploring control methods for systems with shape actuation such as quadrotors with dynamically reconfigurable arms, underwater vehicles with fin actuation, and soft robotics with shape actuation. The end goal of these systems is to realize advances in design, modeling and control of vehicles with actuated shape and actuated and/or dynamic inertia for increased of agility in the sense of faster response times to unexpected obstacles and greater range of response capabilities.

Sparse, distributed and switched sensing

All control systems are subject to some degree of power and processing constraints, which can be limiting for all autonomous vehicles and particularly for small vehicles.  All autonomous vehicles have a variety of onboard sensors and often these sensors provide more sensory information than is required for typical operations.  Our research seeks to build on recent advances in empirical methods for nonlinear control systems to develop tools for assessing which sensory information is the most valuable to a system at any given time for completing a specific task, for maintaining situational awareness, and/or to compensate for sensor failure.

Model-based control and learning

The objective of the work in this proposal is to hybridize model-free Machine Learning (ML) with model-based planning, control, and state estimation to enable active perception for autonomous vehicles operating in complex environments and in complex mission scenarios. Active perception, meaning real-time decision-making and control for task-driven sensing and state estimation, provides a powerful force multiplier for situational awareness at all scales from tactical mapping to basin-scale monitoring, increasing temporal/spatial resolution and providing robustness.

To achieve these capabilities, machine learning methods will provide key insights enabling the high-level autonomy needed to handle complex stochastic environments, whereas model-based algorithms will bring formal robustness needed for reliable mission execution. Our framework decomposes the decision making hierarchically such that machine learning-based and model-based methods are applied where they are most effective via a systematic integration within the hierarchy.

Flight control systems for flexible aircraft

One of the central tradeoffs in aircraft design is the balance between the rigidity of an airframe and the associated structural weight. If an airframe lacks rigidity, i.e. is rather flexible, several issues ensue. Such issues include increased rates of fatigue, degradation of passenger comfort, and adverse interactions between structural deformations and flight dynamic behavior. Advances in actuation, sensing, and control algorithms have set the stage to enable advantageous design of less rigid aircraft by addressing these issues through automatic flight control systems.

The research efforts in this project contribute to that goal with emphasis on the design and analysis of multi-objective control laws with both classical and direct optimization approaches and with validation efforts in both simulation and actively controlled wind tunnel experiments. Some specific topics of interest include the use of Model Predictive Control (MPC) with Light Detection And Ranging (LiDAR) to actively reduce structural loads incurred by atmospheric gusts, measures of robustness for such MPC-based flight control features, and measures of robustness for classically designed multi-objective multi-controller flight control systems.

Control with limited communication

We are interested in question of how much information is necessary and how to provide it for multi-agent systems. We are particularly interested in systems with sensing and communication capabilities that are limited by bandwidth, computational power, range and energy.  Such scenarios are typical of multiple autonomous aircraft in urban settings, multiple autonomous underwater vehicles in shipyards, small autonomous vehicles, and space systems.